- Email: [email protected]
Deep-seated., complex tunnel systems — a hydrological study in a semi-arid catchment, Loess Plateau, China
ELSEVIER
Geomorphology 20 (1997) 255-267
Deep-seated, complex tunnel systems a hydrological study in a semi-arid catchment, Loess Plateau, China T.X. Zhu * Deparm~ent of Geography, Universityof Toronto, St. George Campus, Toronto, ON M5S IA1, Canada
Received 10 June 1995; accepted 25 September 1995
Abstract During 1989-90, detailed monitoring of deep-seated, complex tunnel systems was conducted in Yangdaogou, a subbasin in the Hilly Loess Region, Shanxi, which has a catchment area of 0.203 km 2. It was found that tunnel flow hydrological processes were characterized by a quick response to rainfall, an early flow peak, and a short duration. All the tunnel discharge was derived :from overland flow entering via inlets. However, discharge is not significantly related to rainfall parameters in some tun:ads, which is contrary to observations on surface plots. Field surveys show frequent blockages of tunnels caused by collapses inside the tunnels which could be reopened in subsequent events, as well as the occasional abrupt opening of new inlets. As a result, tunnel discharge was highly erratic in some of the monitored events, Partial damming within the tunnel systems may also be involved. Hence, instability in the tunnel systems is one of the key factors affecting their hydrologic response. During the period of monitoring, fifteen events occurred and on average at least 43% (ranging from 0 to 78%) of the total basin water discharge was routed through four major tunnel systems which account for 90% of the catchment area of all tunnel systems in the basin. The deep-seated tunnel systems in this area seem not to be developed from the micro-pipes close to the surface. Instead, they are most likely formed in some catastrophic storms and expanded in subsequent storms. © 1997 Elsevier Science B.V. Keywords: piping; semi-arid environment; loess;hydrology
1. Introduction The hydrological roles of pipe and tunnel flow in streamflow generation have attracted increasing attention over the past two decades. However, most research has been conducted in humid areas (Morgan, 1972; Gilman and Newson, 1980; Jones, 1981, 1982, 1987; McCaig, 1983; Jones and Crane, i984; Walsh and Howells, 1988), and comparatively few studies have been conducted in s e m i - a d d or a d d areas. Even
* Fax: + 1 416 9786729; E-mail: [email protected]
fewer attempts have been made to monitor the hydrologic significance of deep-seated tunnel systems which refer to those located at depths of over 1 m below the slope surface. Two exceptions exist to this general scarcity. One is the early work conducted by Heede (1971) in Colorado. Despite prolonged monitoring, few records were obtained as flow occurred only during snowmelt. The other is more recent work conducted by Bryan and Harvey (1985) in the Alberta badlands. Two tunnel systems were monitored through three storms. It was observed that about 10% of the total water flow in the study catchment may be routed through tunnel systems.
0169-555X/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S0169-555X(97)00027-5
256
T.X. Zhu / Geomorphology 20 (1997) 255-267
Deep-seated tunnel systems have been reported from many places in semi-arid areas, especially in badlands in Morocco (Imeson et al., 1982), Tunisia (Baillie et al., 1986), Alberta and southern Saskatchewan, Canada (Drew, 1982; Bryan and Harvey, 1985), southeast Spain (Harvey, 1982), Colorado (Heede, 1971), and in loess areas such as the Loess Plateau of China (Fuller, 1922; Thorp, 1936; Chen, 1958), and parts of New Zealand (Cumberland, 1944; Hughes, 1972). The reasons for the comparative lack of detailed field monitoring may be ascribed to the difficulties in the delimitation of individual tunnel systems, the installation of monitoring equipment in the often steep landscapes, as well as the observational problems caused by the shortduration tunnel flow. Nevertheless, to evaluate the hydrological roles of pipe and tunnel flow in a wide area, deep-seated tunnel systems deserve more attention. To serve this purpose, a monitoring program was conducted during 1989 and 1990 in the Yangdaogou, a small subbasin of the Loess Plateau in North China, and all of the tunnel outlets observed to be active were investigated. The preliminary results on tunnel flow hydrology are reported in this paper.
2. The Yangdaogou catchment The Yangdaogou catchment is located about 4 km north of Lishi town, Shanxi Province, China (Fig. 1). It is a first-order drainage basin with an area of 0.203 kl'n 2. The climate is semi-arid warm temperate. Long-term rainfall records at the Shanxi Institute of Soil and Water Conservation, 3 km west of the catchment, indicate that the mean annual rainfall is 499.7 mm, of which over 70% falls within the summer from June to September. Local deposits mainly comprise Quaternary loesses and Tertiary clayey red earth. The loess is believed to be windborne dust derived from central Asia during the Quaternary (Liu, 1964).
3. Description of tunnel systems and tunnel flow monitoring methods The field monitoring area covers the upper and middle parts of the Yangdaogou catchment, with an area of 0.122 km 2. A total of 77 tunnel inlets were found in the Yangdaogou subbasin and 75 are 1o-
cated in the experimental catchment. Both diameter and depth of those tunnel inlets range from less than 0.5 m to more than 20 m, with a mean of 4.8 and 4.99 m, respectively (Fig. 2). In order to delimit the catchment area of the tunnel systems, tunnel networks were first traced with smoke bombs at the beginning of the monitoring period. During the monitoring period, these tracing experiments were also repeated many times in order to detect their temporal changes. The pressure differential between the tunnel inlet and outlet facilitates this method of evaluation. If connectivity is established during any one of the experiments, it is indicated in the topographic map. It was observed that 45 tunnel inlets were connected to one of the 6 outlets although they drained most of the tunnel catchment area. The connected tunnel systems were then surveyed in great detail using a 1 : 1000 topographic map, supplemented by field survey (Table 1). The outlets of tunnels 1, 3, 4 and 6, being located at relatively flat sites, were suitable for the installation of weirs to monitor flow processes. However, the outlets of tunnel 2 and 5 are located on the cliffs, so that two- and three-order flow divisors were installed there, respectively. Although two tubes with a diameter of 10 cm were used to connect the tunnel outlets with the flow divisors, the tunnel flow mixed with trapped air still readily caused the bricks, mortar seal and tubes to burst a few metres away. After frequent repair, I finally had to give up the process of monitoring in those two tunnel systems, although the start and end time of tunnel flow were still recorded during some events. At the exit of the experimental catchment, a concrete flume was constructed by Hamilton (1990) to check the outflow. Additionally, for the purpose of comparison between overland flow and tunnel flow, five surface plots, with areas from 8 to 21,500 m 2, were established in the Yangdaogou subbasin, but only two of the plots were monitored for processes. Owing to the high sediment concentration in the flow, automation of runoff and sediment monitoring was difficult. Thus, stage readings were manually taken every minute throughout the runoff event and sediment samples were taken every three minutes during the first half hour and every six minutes during the second half hour and every twelve minutes thereafter. The stage readings were first converted into discharge using
T.X. Zhu / Geomorphology 20 (1997) 255-267
257
N
CHINA
<3 ~o~
Contour~
Edge of Terrace Tunnel Connectivity . . . . Tunnel Inlet Tlmnel Outlet Surface Plot
U S1
n~e
][
* T1
Raingauge Elevation
\ x
0
30
1008
60
90m
Contour Interval : 5 m
The Yangdaogou Subbasin
Fig. 1. The Yangdaogou experimental catchment, tunnel connecfivities and measurement sites.
T.X. Zhu/ Geomorphology20 (1997) 255-267
258
20~ 18" 1614~ 12= 1024
~" 8.
I
6-
(m)
0 ,v Io Diameter (m)
21
late July in 1989 throughout the rainy season in 1990. The annual precipitation was 340 m m in 1989 and 623 m m in 1990 (Fig. 3a), w h i c h represents dry and wet years with return periods o f 5 to 10 years, respectively (Table 2). A total o f fifteen storms were m o n i t o r e d and are characterized by a short duration and an early peak. B y c o m p a r i n g the f r e q u e n c y o f m a x i m u m 3 0 - m i n rainfall intensity b e t w e e n the m o n itored storms and the l o n g - t e r m rainfall records ( 1 9 5 5 - 1 9 9 0 ) o f the Shanxi Institute of Soil and W a t e r Conservation, we found that the m o n i t o r e d storms c o v e r almost the full range o f runoff-generation storms in this area except the e x t r e m e l y h e a v y ones (Fig. 3b).
24
Fig. 2. Frequency distribution of tunnel inlet size.
4. Results and discussion the f o r m u l a e d e v e l o p e d and tested by Z e n g (1983) and then s e d i m e n t discharges in the f l o w w e r e further excluded. The field m o n i t o r i n g p r o g r a m was c o n d u c t e d f r o m
4.1. Tunnel flow hydrological responses (a) Tunnel flow start time. Tunnel f l o w start time is here defined as the lag time b e t w e e n the onset o f
Table 1 Major characteristics of the tunnel systems in Yangdaogou subbasin Tunnel system
Morphometry No. of Minimum inlets tunnel length (m)
Physical characteristics Mean slope
Outlet area (m 2)
Minimum catchment area (m2)
1
4-5
2
0.35
0.13
2,410
2
1
6
1.08
0.68
2,050
8-12
132
0.57
0.13
12,700
5-9
45
0.72
0.36
3,285
3
44
0.46
0.3
3,210
254
0.36
3.10
25,600
5
6
24-25
Developed on an old landslide and the bottom of the tunnel is the contact between loess and red earth. It drains water from the gentle slopes covered with dense shrub (Caragana). Formed within loess deposits. Two layers of tunnel conduits are connecting to the outlet but the upper one is abandoned. Tunnel drains water from cultivated slopelands. Developed within the loess and the outlet is located just above the gully bottom land. Three series of inlets ranked by size up the valley slope. Tunnel catchment consists of a large proportion of terrace lands. Developed within the loess and the outlet; is also located just above the gully bottom land. Tunnel catchment underwent a great change over the monitoring period. The outlet is hanging on the cliff. Though the uppermost tunnel inlet is adjacent to the above terrace land there is no indication that overland flow enters the tunnel system. Thus the main drainage area is confmed to the vailey slope. The largest tunnel in the basin. Red earth is scattered along the gully bottom of the catchment. In the 1990 rainy season, a new inlet leading to 40 m or so tunnel conduit was formed during one single storm.
T.X. Zhu / Geomorphology 20 (1997) 255-267 700
259
(a) _.=
600
A :500 E
g
,~
f
m
E O 200
100
o
50
I
I
I
I
I
I
100
150
200
250
300
350
Day of Year
50
40
(b) ¸
[] Long Term • Mon tored Period
A
o
30
O"
u.
20
10
0-9.9
10-19.9
20-29.9
30-39.9
40-49.9
Max 30-min Intensity (ram/h)
Fig. 3. (a) Cumulative rainfall in the Yangdaogou in 1989 and 1990. (b) Frequency distribution of maximum 30-min rainfall intensity for runoff-generation storms in a long-term and the monitored period.
Table 2 Annual precipitation with various return periods Category
Wet year
Return periods Annual precipitation (mm)
200 908
After Zhang et al. (1992).
100 865
50 807
20 733
10 671
5 601
Average year
Dry year
2 483
5 399
10 334
20 299
50 240
T.X. Zhu / Geomorphology 20 (1997) 255-267
260
rainfall and the emergence of flow at the tunnel outlet. Tunnel flows occurred in twelve of the fifteen monitored storms. Overall, their response to rainfall was very fast, with the start lag times ranging from 1 to 67 min (Table 3). No distinctive difference exists between tunnel flow start time and overland flow start time, which implies that tunnel flow velocity is very fast. Although no measurements were made, we could roughly estimate it. The shortest start time for tunnel 6 was 5 min. Even if we assume that overland flow was initiated immediately and entered the last tunnel inlet after the rainfall start, the tunnel flow still had an average velocity of 0.26 m s-]. It is noted that the tunnel length of this section was actually measured by the author passing through it. (b) Tunnel flow peaks A total of 42 discharge peaks were observed from 35 complete process records at four tunnels (numbers 1, 3, 4, 6). Most tunnel hydrographs only have a single peak, which occurs within half an hour after the start of tunnel flow. The number of peaks seems to be determined by rainfall characteristics and is not affected by tunnel complexity. But exceptions did exist. For example, the storm of I 1 July 1990 was characterized by a single rainfall peak. Likewise, one discharge peak occurred in tunnels 1, 4 and 6. How-
2
,~ ~'1.5
"~'E !go,
Rainfall
I o~
J
II
I
I
I'
I
200
...;;
150 / t I t
100 u
?a O
50
i I
+
Tunnel Hydrograph
------T4
,~
0
"
60O 500 400 300 •~ 200 ._ 100 tn 0
I
J
Flume Hydrograph
L 20
40
60
80
100
120
Time (min)
Fig. 4. Rainfall, tunnel and basin outflow hydrographs of the August 11, 1990 storm.
ever, two tunnel flow peaks were produced in tunnel 3 (Fig. 4). The reason is still unclear. One possible explanation is that this is due to the runoff generation zonation. In light or medium storms, runoff in
Table 3 Tunnel and overland flow start time Date
Precipitation
Antecedent rainfall
First 10-min intensity
Runoff start time (min)
(mm)
(mm) (5 days)
(mm/min)
TI
T2
T3
T4
T5
T6
S1
F
6 August 89 10 August 89 15 August 89 16 August 89
24.8 13.1 28.7 29.1
0 24.8 13.1 28.7
0.99 0.37 0.65 0.15
10 N 21.4 N
16 N 4.2 N
4 N 18 N
8.7 N 17.5 N
13 N M N
5 N 20 N
M M M M
13 25 15 10
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
8.4 29.5 30.5 49.5 1.6 46.9 33.3 53.1
0.04 1.19 1.8 0.16 0.42 1.05 0.09 0.4
N 12 15 49 67 19 M M
N 15 8 29 40 1.0 M 22
N 13 11 59 M 9 M 30
N 12 14 12 30 11 M 25
N M 8 29 20 8 M M
N 23 17 15 10 17 18 12
110 13 12 53 73 6 M 15
M 3 11 27 12 7 6 17
11 August 90 13 August 90 28 August 90
20.2 35.4 53
3.9 24.1 0
0+63 09 1.4
20 M 13
17 29 15
23 M 23
20 13 17
M M M
20 15 15
18.3 21 21 M M 20
T = tunnel system; S = surface plot; F = experimental basin flume; M = missed events; N = no runoff occurs.
T.X. Zhu / Geomorphology 20 (1997) 255-267
the tunnel catchment is mainly generated from the cultivated slopes and the steeper valley side slopes and can be quickly directed into the tunnel system. However, in heavy storms such as the above-mentioned one, a large amount of runoff can also be generated from the e:~tensive terrace lands and takes a considerably longer time to reach the tunnel inlets than the former owing to the relatively remote locations and gentle slopes. Therefore, it formed a second discharge peak for the same rainfall peak. The uneven peaks in the flume hydrograph in Fig. 4 imply the slight difference in peak time among different tunnels. (c) Tunnel flow duration. In light and medium storms, tunnel flow duration shows only limited differences among the tunnels, and is more or less comparable to the effective rainfall duration (Table 4). Here, effective r,'dnfall is defined as the rainfall with an intensity of over 0.2 m m / m i n (Zhu et al., 1995). However, in the heavy storms, such as on 11 July 1990, 26 July 1990 and on 28 August 1990, tunnels 3 and 4 had a considerably longer flow duration than the others. This is because the runoff could be generated from the terrace lands in heavy storms and it took quite a long time to enter the
261
tunnel networks for tunnels 3 and 4, whereas the tunnel flow duration for the remaining ones was still quite similar to that in light and medium storms, owing either to relatively shorter distances between terrace land and nearest inlets or to higher slope gradients in the tunnel catchments.
4.2. Impacts of instability within tunnel systems on tunnel flow hydrology Deep-seated tunnel systems in this area are characterized by great instability. Collapses within tunnel systems are very common. Small-scale collapses may only cause oscillation of sediment concentrations in the tunnel flow and no effects on tunnel flow hydrology. However, large collapses could exert profound impacts on tunnel flow hydrology. If the collapses are extremely large or associated with surface depression or sediment deposition, tunnel systems could be totally blocked. Rapid tunnel flow could, in turn, reopen the blocked tunnel systems. Such temporal shift of tunnel systems can be detected using smoke bombs before and after storm events. In the 1989 rainy season, the outlet of tunnel 1 was connected to two series of inlets, but the major branch was blocked
Table 4 Tunnel and overland flow duration Date
Precipitation
Antecedent rainfall
Rainfall duration
Rainfall with 1 > 0.15 m m / m i n duration
Runoff duration (min)
(mm)
(mm)(5 days)
(rain)
(min)
T1
T2
T3
T4
T5
T6
S1
F
6August89 10August 89 15August89 16 A u g u ~ 89
24.8 13.1 28.7 29.1
0 24.8 13.1 28.7
45 35 170 535
45 35 60 10
25 N 21 N
29 N 41 N
58 N 37 N
32 N 45 N
31 N M N
41 N 40 N
M M M M
61 25 165
445
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
8.4 29.5 30.5 49.5 1.6 46.9 33.3 53.1
450 8 132 110 67 185 220 85
0 8 42 110 30 70 60 50
N 11 37 51 31 60 M M
N 10 9 41 14 46 M 25
N
N
7 96 71 M 103 M 26
6 51 55 28 120 M 29
N M M M M 50 M M
N 9 21 57 45 48 27 29
7 17 27 56 51 108 M 34
M 34 82 96 60 105 141 60
llAugust90 13 August90 28August90
20.2 35.3 53
3.9 24.1 0
83 914 255
30 54 140
24 M 62
21 40 42
32 M 92
28 77 83
M M M
39 52 41
36 38 42
53 M 63
T = tunnel system; S = surface plot; F = experimental basin outflow flume; M = missed events; N = no runoff occurs.
T.X. Zhu / Geomorphology 20 (1997) 255-267
262
in the 1990 rainy season. This led to a great disparity in tunnel flow discharge between 1989 and 1990 (Fig. 5). In tunnel 3, the southern branch, consisting of four tunnel inlets, was blocked throughout the 1989 and 1990 rainy seasons, but it was reopened in 1992 (DiCenzo, pers. commun., 1995). The middle branch was also blocked during the storm of 11 August 1990 and reopened by the storm of 28 Au-
60 E ~.=_ 40
1--
gust 1990, which caused the tunnel flow discharge of the former storm to be disproportionally low. In tunnel 4, three connecting tunnel inlets, about 30 m south of the tunnel outlet, were abruptly joined into tunnel 4 during the storm of 13 August 1990. Those newly joined tunnel inlets subsequently added a large amount of runoff to the tunnel system generated from slope land and terrace land.
.Pr~.tie0.~mm~
.
=
1
20
o
75 50
N
I~ N
o 100 75
Tunnel3
25
0
1°t ~ O0
Tunnel4
o
300 loo
250
Tunnel6
2OO 150
N
50
o
Date M: Missed events N: No runoff
Fig. 5. Water discharges at tunnels 1, 3, 4 and 6 in 1989 and 1990.
1
I1
263
T.X. Zhu / Geomorphology 20 (1997) 255-267
The most significant event which I observed during the two consecutive rainy seasons was the abrupt initiation of one tunnel inlet on 13 August 1990. The inlet, with a diameter of 1.5 m and depth of 1.9 m, was developed in the middle of a road, located on the upper drainage boundary of tunnel 6. Smoke bomb tests indicated that it was connected to an inlet of tunnel 6 about 40 m away! The runoff from the village and the neighbouring subbasin, which used to flow into another ba:~in via the excavated road, was redirected into tunnel 6 through the inlet and conduit of the newly developed tunnel. This led to the discharge of tunnel 6 being unusually high during the storm of 13 August 1990 (Fig. 5). After that storm, the inlet was filled in by the villagers, since it hindered traffic. As a result, discharge at tunnel 6 returned to normal in the subsequent storm of 28 August 1990. Totally blocked tunnel branches can be readily detected with smoke bombs and thereby their impacts on tunnel flow hydrology can be explicitly evaluated. However, in most cases, the tunnel systems may not be totally blocked but partially dammed or blocked first and reopened later during the same storm. In these situations, smoke bombs are useless and it is extremely dangerous to investigate by crawling into the tunnel systems after storms. Thus normally no direct evidence is available. However, in July of 1989, I did manage to pass through the last section of tunnel 6, a 76 m long tunnel conduit, and found a bridge-like constriction inside. Apparently, it used to be a dam caused by a collapse and sometime later tunnel flow penetrated the dam and formed the opening under the 'bridge'. Here, in contrast to the intended objective of this section, I use the monitored tunnel flow hydrologic processes to identify the possible partial damming during the event. Owing to the lack of direct evidence, the results presented here must be considered tentative and to be examined further in the future. In the storm of 26 July 1990 (Fig. 6), tunnel flow processes in all tunnels except tunnel 4 were characterized by an early discharge peak, which was caused by an immediate intensity peak after rainfall onset. However, discharge at tunnel 4 was very low in the first hour and peak discharge did not occur until 77 min after rainfall onset. After the peak, the discharge sharply dropped to a very low level and lasted for another 40
1.5 Rainfall
_= E
.o 0
60 I I I t I l t"
40 30
" ° "T3 -----T4 -- -- -- T6
Tunnel Hydrograph
,Oo 200| $ lso~
/
~
FlumeHydrogmph
=.;e'oto 0
I 20
40
60
80
I I 100 120 140 160 180 200
Tlrne (mln)
Fig. 6. Rainfall, tunnel and basin outflowhydrographsof the July 26, 1990 storm.
min or so. The total discharge appeared to be normal and the absence of the first peak probably resulted from partial damming, which was flushed away later by accumulated water inside the tunnel. In contrast, in tunnel 6, after the first discharge peak, tunnel flow simply stopped. It was unlikely that no runoff had been generated by the second rainfall peak from the tunnel catchment, the largest one in the basin. The total discharge from tunnel 6 during this storm was also quite low. Twelve hours later, another storm occurred, with a rainfall of 19.8 mm. Flow in tunnel 6 started 18 min after rainfall onset and the discharge was so high that the trapezoidal weir shifted. Accordingly, no water discharge data were collected, although sediment samples were still taken throughout the event. Actually, the mean rainfall intensity for the first 20 min (0.09 mm/min) was very low in this storm. Though the antecedent soil moisture was very high, it is unlikely to have produced such a high flow if no water had been trapped by damming during the previous storm. Two rainfall peaks produced four discharge peaks in tunnel 3 (Fig. 6; similar to that shown in Fig. 4), may have been caused by runoffgeneration zonation as well.
264
T.X. Zhu / Geomorphology 20 (1997) 255-267
Table 5 R 2 values in linear regression analyses between runoff discharge and effective rainfall ( l > 0.20 m m / m i n ) Overland flow
Tunnel flow
Plot No.
R 2 (including all events)
Tunnel No.
R 2 (including all events)
R 2 (discarding events affected by instability)
S-1 (zone 1) S-2 (zones l a n d 2) S-3 (zones 1, 2 a n d 3)
0.642 (30) 0.535 (34) 0.617(40)
T1 T3 T4 T6
0.075 0.691 0.257 0.125
0.622(10) 0.713 (11) 0.542 (9) 0.301 (11)
(12) (12) (11) (13)
Note: Values in parentheses represent the number of events.
To further evaluate the impact of instability on tunnel flow discharge, I compared it with overland flow discharge. Owing to the limited number of events that were monitored on the surface plots during our monitoring periods, I have used the data collected by the Shanxi Institute of Soil and Water Conservation during the period of 1963-1968 from the Yangdaogou subbasin (SISWC, 1982). Three surface plots with areas of 400, 1855 and 4167 m 2 were selected in the subbasin comprising upper slope, lower slope, and combined slope, respectively. It
was found that a good correlation exists between runoff discharge and rainfall with an intensity of more than 0.2 m m / m i n for all three surface plots (Table 5). For tunnel systems, such a good correlation could only be~found for tunnel 3. However, if we disregard thoseevents affected by tunnel instability, identified above, correlation coefficients are improved for all tunnels, especially for tunnel 1 and 4. The still poor correlation for tunnel 6 may be ascribed to unidentified tunnel instabilities within this large and complex system.
Table 6 Contribution of tunnel flow to the experimental catchment outflow Date
Rainfall (mm)
Effective rainfall (mm)
Tunnel flow a (m 3)
6 August 89 10 August 89 15 August 89 16 August 89
24.8 13.1 28.7 29.1
24.8 13.1 22 8.9
142.2 0 194.7 0
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
5.2 9.5 35.4 28.5 15.5 28.4 16.6 14.2
11 August 90 13 August 90 28 August 90
20.2 35.4 53
Total
390.1
Tunnel flow b (m 3)
Outflow (m 3)
Contributions a (%)
Contributions b (%)
224.0 0 221.8 0
280.7 36.6 330.7 57.3
50.7 0 58.9 0
79.8 0 67.1 0
0 6.1 190.0 141.1 64.3 147.5 0 80.6
0 19.3 218.2 153.6 104.5 164.6 82.5 85.8
120.5 85.6 940.6 283.5 164.5 360.4 332.8 104.5
0 7.2 20.2 49.8 39.1 40.9 0 77.1
0 22.5 23.2 54.2 63.5 45.7 25.0 82.1
18.1 22.9 44.7
20.0 353.8 189.9
23.9 486.6 219 8
79.8 389.2
25.0 48.8
29.9 56.5
307.8
1530.2
1871.6
3566.7
42.9
52.5
Notes: a Estimated from Method 1. b Estimated from Method 2.
T.X. Zhu / Geomorphology 20 (1997) 255-267
4.3. Hydrological significance of tunnel flow Tunnel flow discharge data are only available from tunnels 1, 3, 4 and 6, but they account for about 90% of the catchment area of all tunnel systems in the basin. Here, two methods are used to evaluate the contribution of tunnel flow to the basin outflow. In Method 1, only those monitored events and tunnel systems are considered which could represent the minimum contribution of tunnel flow. In Method 2, catchmen)I area and effective rainfall are used to interpolate the discharges of tunnels 2 and 5, as well as those missed events for the other four tunnels. The results are listed in Table 6. Overall, the total tunnel flow contribution to basin outflow is 42.9-52.5%, which is slightly higher than the proportion of tunnel catchments to total experimental area, 40% or so. Great interstorrn variations also exist. In the light storms with low intensities such as on 10 August 1989, 16 August 1990 and 6 July 1990, no tunnel flow was observed though overland flow was generated. Field investigations showed that overland flow generated within the tunnel catchment area x~as reduced by collapse debris or the surrounding materials of the tunnel conduits while
265
flowing into tunnel systems via inlets. Such reduction was maximized at the beginning of the rainy season and increased with tunnel length. Because we did not monitor the beginning of the 1989 rainy season, I analyze here the temporal variation of tunnel flow discharge only for the 1990 rainy season. The first runoff-generation storm occurred on July 6, with a total of 21 mm precipitation, and 120 m 3 runoff passing through the low flume, but no tunnel flow emerged from any tunnel outlets. On the following day, a total of 9.5 m m rain fell within 8 min and tunnel flow occurred in all tunnel systems except tunnel 6. Tunnel 1, with the blockage of a longer branch (70 m or so) and the opening of a shorter one (2 m or so), had an even higher runoff coefficient than the whole experimental catchment (Fig. 7). However, for tunnel 3, with an at least 45 m long conduit from the last inlet to outlet, the nmoff coefficient was still very low. No tunnel flow was observed from the outlet of tunnel 6, with a 76 m long conduit from last inlet to outlet. Even in the storm of 11 July, the third one of the rainy season, tunnel 6 contributed only 62.3 m 3 or 7.3% runoff to the experimental subbasin outflow. This was still disproportionally smaller than its catchment proportion to
25
llil T1
20
l
liT3
lilT4 •T6 • Subbas
A
15 U
E
u ~ lO e,-
6-Jul
7-Jul
11-Jul
13-Jul
Date
Fig. 7. Temporal variations in runoff coefficients of tunnels and experimental catchment at the beginning of the 1990 rainy season.
266
T.X. Zhu / Geomorphology 20 (1997) 255-267
the subbasin, 21%, whereas runoff coefficients at other tunnels were more or less comparable to the whole experimental catchment. Hamilton (1990) has also stated " f o r some tunnel outlets there was no evidence of flow during the first two events of the year [1988] whereas major discharges were directly observed from these tunnels during the third events of the rainy season." From Table 6, it can be seen that the tunnel flow contributions to the subbasin outflow are often significantly high from medium storms such as on 6 and 15 August 1989, and on 22 and 30 July 1990. This may be due to several reasons. First, after the tunnel was wetted at the beginning of the rainy season, the thick tunnel roof prevents evaporation from wetted tunnel conduits. Second, after sheet or rill flows enter the tunnel systems, infiltration losses should be reduced owing to the small contact area. Third, particle size analyses indicated that all six tunnel floors were composed of weak permeable layers with higher clay contents than the surface (Luk and Zhu, 1995), which could further reduce infiltration losses. In very heavy storms, such as those on 11, 26 and 28 July 1990, tunnel flow contributions to the outflow fall off again and are more or less comparable to the proportion of tunnel catchment areas to the basin. This is because the overland flow is generated throughout the basin and the difference in infiltration losses between tunnel systems and untunnelled surface areas is considerably reduced. Additionally, heavy storms may result in active collapses in the tunnel systems, which could trap some tunnel flow.
5. Conclusion and implications Deep-seated tunnel systems in this area act as conduits for overland flow. Soil throughflow is too shallow to reach the tunnel systems during storm events. Field experiments and investigations indicate that the wetting depth is usually less than 30 cm and rarely exceeds 50 cm during the storms (Li Gang, 1992), whereas almost all the tunnel systems are located much below this. The ground water is too deep to recharge the tunnel flow; although some tunnel inlets are 20 m deep they are still not deep enough to be close to the ground water table.
Tunnel flow does not simply mirror overland flow. First, high levels of absorption of water at the beginning of the rainy season and the high efficiency of flow delivery later create significant differences in the temporal variation of flow discharge over the season. Second, instability within the tunnel systems could make tunnel flow discharge or processes highly erratic within storms. Third, some inlets not connected to any outlets still drain water from the catchment. The trapped water is not only absorbed by surrounding materials but may also recharge the ground water, which could be a very important process in this area. Field rainfall simulation experiments conducted in this area showed that micro-pipes could also be developed close to the surface. Initiation of those shallow pipes is closely related to rill development (Bryan et al., 1978; Yair et al., 1980; Wang, 1991). However, in this area, they seem unlikely to develop into deep-seated tunnels. Instead, the huge tunnel systems are more likely to be initiated in catastrophic storms, although the subsequent storms do expand them. One tunnel inlet leading to about 40 m of tunnel conduits, was developed in a single storm on the 13th July 1990. Another example was reported from Wudu, Gansu Province in the western Loess Plateau. A total of 123 tunnel inlets were developed within an area of 0.2 km 2 during a single storm with a precipitation of 37 mm and duration of 115 min in 1962 and the volume of developed tunnel totalled around 1000 m 3 (Wang, 1989). These eroded sediments mixed with runoff and developed into so-called hyperconcentrated flow and were flushed out to stream channels via existing or newly developed conduits (to be discussed in a separate paper, Luk and Zhu, 1995).
Acknowledgements This study was funded by the Canadian International Development Agency. The author is grateful to Profs. S.H. Luk, Chen Yongzong, Cai Qiangguo and Zeng Boqing for their valuable advice during the setting up of monitoring equipment. I am also very much indebted to both Canadian and Chinese research teams for their support in the field. The manuscript significantly benefited from the com-
T.X. Zhu / Geomorphology 20 (1997) 255-267 m e n t s a n d s u g g e s t i o n s o f Prof. R. B r y a n a n d t w o a n o n y m o u s referees.
References Baillie, I.C., Fanlkner, P.H, Espin, G.D., Levett, M.J., Nicholson, B., 1986. Problems of protection against piping and surface erosion in central Tuni,;ia. Environ, Conserv. 13, 27-32. Bryan, R.B., Harvey, L.E., 1985. Observation on the geomorphologic significance of tunnel erosion in a semi-add ephemeral drainage system. Geogr. Ann. 67 (A), 257-272. Bryan, R.B., Yair, A., Hodges, W.K., 1978. Factors controlling the initiation of runoff and piping in Dinosaur Provincial Park badland, Alberta, Canada. Z. Geomorphol. Suppl. 29, 151-168. Chen, Y.Z., 1958. Tunnel formation and development in the Loess Plateau. Geography Notes, 2:25-30 (in Chinese). Cumberland, K.B., 1944. Soil Erosion in New Zealand. Soil Conservation and River Control Council, Wellington, 288 pp. Drew, D.P., 1982. Piping in the big muddy badlands, southern Saskatchewan, Canada. In: Bryan, R., Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 293-304. Fuller, M.L., 1922. Some unusual erosion features in the loess of China. Geogr. Rev. 12, 570-584. Gilman, K., Newson, M.D., 1980. Soil Pipes and Pipeflow--a Hydrological Study in Upland Wales. British Geomorphological Research Group Monogr. 1, GeoBooks, Norwich, 114 pp. Hamilton, H., 1990. Field Examination of Soil Erosion and Losses of Nitrogen and Phosphorus from the Agricultural Watershed, I. The Loess Plateau, People's Republic of China. Unpublished Ph.D. thesis, University of Toronto, 205 pp. Harvey, A., 1982. The vale of piping in the development of badlands and gully systems in southeast Spain. In: Bryan, R., Yair, A. (Eds.), Badl~nd Geomorphology and Piping. GeoBook, Norwich, pp. 317-335. Heede, B.H., 1971. Characteristics and processes of soil piping in gullies. U.S. Department of Agriculture, Forest Service Research Paper, Rm 68, 15 pp. Hughes, P.J., 1972. Slope aspect and tunnel erosion in the loess of Banks Peninsula, New Zealand. J. Hydrol. (New Zealand) 11, 94-98. Imeson, A.C., Kwaad, lz.J.P.M., Verstraten, J.M., 1982. The relationship of soil physical and chemical properties to the development of badlands in Morocco. In: Bryan, R. Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 47-70. Jones, J.A.A., 1981. The Nature of Soil Piping--a Review of Research. British Geomorphological Research Group Monogr. 3, GeoBooks, Norwich, 301 pp. Jones, J.A.A., 1982. Experimental studies of pipe hydrology. In:
267
Bryan, R. Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 355-370. Jones, J.A.A., 1987. The effects of soil piping on contributing areas and erosion patterns. Earth Surf. Process. Landf. 12, 229-248. Jones, J.A.A., Crane, F.G., 1984. Pipeflow and pipe erosion in the Maesnant experimental catchment. In: Burt, T.P., Wailing, D.E. (Eds.), Catchment Experiments in Fluvial Geomorphology. GeoBooks, Norwich, pp. 55-72. Li Gang, 1992. Experimental Study on Rainfall Infiltration--one of the Fundamentals of Hillslope Amelioration in Loess Plateau. Unpublished Ph.D. thesis, Institute of Geography, Chinese Academy of Sciences (in Chinese with English abstract). Liu, D.S., 1964. Loess Deposits in the Middle Reaches of Yellow River. Science Press, Beijing, 234 pp. (in Chinese). Luk, S.H., Zhu, T.X., 1995. Tunnel erosion and sediment production in the hilly loess region, North China. (in preparation). McCaig, M., 1983. Contributions to storm quickflow in a small headwater catchment: the role of natural pipes and soil macropores. Earth Surf. Process. Landf. 8, 239-252. Morgan, R.P.C., 1972. Observation on factors affecting the behavior of a first-order stream. Trans. Inst. Br. Geogr. 56, 171-185. SISWC (Shanxi Institute of Soil and Water Conservation), 1982. Rainfall, runoff and sediment data (1955-1981). Unpublished (in Chinese). Thorp, J., 1936. Geography of the soils of China. National Geological Survey of China, Nanking. Waish, R.P.D., Howells, K.A., 1988. Soil pipes and their role in runoff generation and chemical denudation in a humid tropical catchment in Dominica. Earth Surf. Process. Landf. 13, 9-17. Wang, B.K., 1989. Factors causing tunnel erosion. Acta of Soil and Water Conservation, 3(3): 84-90 (in Chinese). Wang, G.P., 1991. Modeling Soil Erosion from Hillslope in the Hilly Loess Region, China. Unpublished M.Sc. thesis, University of Toronto, 117 pp. Yair, A., Bryan, R.B., Lavee, H., Adar, E., 1980. Runoff and erosion processes and rates in the Zin Valley badlands, Northern Negev, Israel. Earth Surf. Process. 5, 205-225. Zeng, B.Q., 1983. Methods for soil and water losses monitoring in the small catchment. Soil and Water Conservation of Shanxi, 2:11-19 (in Chinese). Zhang, Z.G., Zen, B.Q., Jia, Z.J., 1992. Precipitation characteristics in Wangjiagou basin, western Shanxi. In: Wang, F.T. (Editor), Soil Erosion Management and Application of Geographical Information Systems in the Loess Plateau region, western Shanxi. pp. 1-12 (in Chinese). Zhu, T.X., Cai, Q.G., Zeng, B.Q., 1995. Runoff generation in a semi-arid agricultural catchment: field and experimental studies. Submitted to Journal of Hydrology (under review).
Geomorphology 20 (1997) 255-267
Deep-seated, complex tunnel systems a hydrological study in a semi-arid catchment, Loess Plateau, China T.X. Zhu * Deparm~ent of Geography, Universityof Toronto, St. George Campus, Toronto, ON M5S IA1, Canada
Received 10 June 1995; accepted 25 September 1995
Abstract During 1989-90, detailed monitoring of deep-seated, complex tunnel systems was conducted in Yangdaogou, a subbasin in the Hilly Loess Region, Shanxi, which has a catchment area of 0.203 km 2. It was found that tunnel flow hydrological processes were characterized by a quick response to rainfall, an early flow peak, and a short duration. All the tunnel discharge was derived :from overland flow entering via inlets. However, discharge is not significantly related to rainfall parameters in some tun:ads, which is contrary to observations on surface plots. Field surveys show frequent blockages of tunnels caused by collapses inside the tunnels which could be reopened in subsequent events, as well as the occasional abrupt opening of new inlets. As a result, tunnel discharge was highly erratic in some of the monitored events, Partial damming within the tunnel systems may also be involved. Hence, instability in the tunnel systems is one of the key factors affecting their hydrologic response. During the period of monitoring, fifteen events occurred and on average at least 43% (ranging from 0 to 78%) of the total basin water discharge was routed through four major tunnel systems which account for 90% of the catchment area of all tunnel systems in the basin. The deep-seated tunnel systems in this area seem not to be developed from the micro-pipes close to the surface. Instead, they are most likely formed in some catastrophic storms and expanded in subsequent storms. © 1997 Elsevier Science B.V. Keywords: piping; semi-arid environment; loess;hydrology
1. Introduction The hydrological roles of pipe and tunnel flow in streamflow generation have attracted increasing attention over the past two decades. However, most research has been conducted in humid areas (Morgan, 1972; Gilman and Newson, 1980; Jones, 1981, 1982, 1987; McCaig, 1983; Jones and Crane, i984; Walsh and Howells, 1988), and comparatively few studies have been conducted in s e m i - a d d or a d d areas. Even
* Fax: + 1 416 9786729; E-mail: [email protected]
fewer attempts have been made to monitor the hydrologic significance of deep-seated tunnel systems which refer to those located at depths of over 1 m below the slope surface. Two exceptions exist to this general scarcity. One is the early work conducted by Heede (1971) in Colorado. Despite prolonged monitoring, few records were obtained as flow occurred only during snowmelt. The other is more recent work conducted by Bryan and Harvey (1985) in the Alberta badlands. Two tunnel systems were monitored through three storms. It was observed that about 10% of the total water flow in the study catchment may be routed through tunnel systems.
0169-555X/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S0169-555X(97)00027-5
256
T.X. Zhu / Geomorphology 20 (1997) 255-267
Deep-seated tunnel systems have been reported from many places in semi-arid areas, especially in badlands in Morocco (Imeson et al., 1982), Tunisia (Baillie et al., 1986), Alberta and southern Saskatchewan, Canada (Drew, 1982; Bryan and Harvey, 1985), southeast Spain (Harvey, 1982), Colorado (Heede, 1971), and in loess areas such as the Loess Plateau of China (Fuller, 1922; Thorp, 1936; Chen, 1958), and parts of New Zealand (Cumberland, 1944; Hughes, 1972). The reasons for the comparative lack of detailed field monitoring may be ascribed to the difficulties in the delimitation of individual tunnel systems, the installation of monitoring equipment in the often steep landscapes, as well as the observational problems caused by the shortduration tunnel flow. Nevertheless, to evaluate the hydrological roles of pipe and tunnel flow in a wide area, deep-seated tunnel systems deserve more attention. To serve this purpose, a monitoring program was conducted during 1989 and 1990 in the Yangdaogou, a small subbasin of the Loess Plateau in North China, and all of the tunnel outlets observed to be active were investigated. The preliminary results on tunnel flow hydrology are reported in this paper.
2. The Yangdaogou catchment The Yangdaogou catchment is located about 4 km north of Lishi town, Shanxi Province, China (Fig. 1). It is a first-order drainage basin with an area of 0.203 kl'n 2. The climate is semi-arid warm temperate. Long-term rainfall records at the Shanxi Institute of Soil and Water Conservation, 3 km west of the catchment, indicate that the mean annual rainfall is 499.7 mm, of which over 70% falls within the summer from June to September. Local deposits mainly comprise Quaternary loesses and Tertiary clayey red earth. The loess is believed to be windborne dust derived from central Asia during the Quaternary (Liu, 1964).
3. Description of tunnel systems and tunnel flow monitoring methods The field monitoring area covers the upper and middle parts of the Yangdaogou catchment, with an area of 0.122 km 2. A total of 77 tunnel inlets were found in the Yangdaogou subbasin and 75 are 1o-
cated in the experimental catchment. Both diameter and depth of those tunnel inlets range from less than 0.5 m to more than 20 m, with a mean of 4.8 and 4.99 m, respectively (Fig. 2). In order to delimit the catchment area of the tunnel systems, tunnel networks were first traced with smoke bombs at the beginning of the monitoring period. During the monitoring period, these tracing experiments were also repeated many times in order to detect their temporal changes. The pressure differential between the tunnel inlet and outlet facilitates this method of evaluation. If connectivity is established during any one of the experiments, it is indicated in the topographic map. It was observed that 45 tunnel inlets were connected to one of the 6 outlets although they drained most of the tunnel catchment area. The connected tunnel systems were then surveyed in great detail using a 1 : 1000 topographic map, supplemented by field survey (Table 1). The outlets of tunnels 1, 3, 4 and 6, being located at relatively flat sites, were suitable for the installation of weirs to monitor flow processes. However, the outlets of tunnel 2 and 5 are located on the cliffs, so that two- and three-order flow divisors were installed there, respectively. Although two tubes with a diameter of 10 cm were used to connect the tunnel outlets with the flow divisors, the tunnel flow mixed with trapped air still readily caused the bricks, mortar seal and tubes to burst a few metres away. After frequent repair, I finally had to give up the process of monitoring in those two tunnel systems, although the start and end time of tunnel flow were still recorded during some events. At the exit of the experimental catchment, a concrete flume was constructed by Hamilton (1990) to check the outflow. Additionally, for the purpose of comparison between overland flow and tunnel flow, five surface plots, with areas from 8 to 21,500 m 2, were established in the Yangdaogou subbasin, but only two of the plots were monitored for processes. Owing to the high sediment concentration in the flow, automation of runoff and sediment monitoring was difficult. Thus, stage readings were manually taken every minute throughout the runoff event and sediment samples were taken every three minutes during the first half hour and every six minutes during the second half hour and every twelve minutes thereafter. The stage readings were first converted into discharge using
T.X. Zhu / Geomorphology 20 (1997) 255-267
257
N
CHINA
<3 ~o~
Contour~
Edge of Terrace Tunnel Connectivity . . . . Tunnel Inlet Tlmnel Outlet Surface Plot
U S1
n~e
][
* T1
Raingauge Elevation
\ x
0
30
1008
60
90m
Contour Interval : 5 m
The Yangdaogou Subbasin
Fig. 1. The Yangdaogou experimental catchment, tunnel connecfivities and measurement sites.
T.X. Zhu/ Geomorphology20 (1997) 255-267
258
20~ 18" 1614~ 12= 1024
~" 8.
I
6-
(m)
0 ,v Io Diameter (m)
21
late July in 1989 throughout the rainy season in 1990. The annual precipitation was 340 m m in 1989 and 623 m m in 1990 (Fig. 3a), w h i c h represents dry and wet years with return periods o f 5 to 10 years, respectively (Table 2). A total o f fifteen storms were m o n i t o r e d and are characterized by a short duration and an early peak. B y c o m p a r i n g the f r e q u e n c y o f m a x i m u m 3 0 - m i n rainfall intensity b e t w e e n the m o n itored storms and the l o n g - t e r m rainfall records ( 1 9 5 5 - 1 9 9 0 ) o f the Shanxi Institute of Soil and W a t e r Conservation, we found that the m o n i t o r e d storms c o v e r almost the full range o f runoff-generation storms in this area except the e x t r e m e l y h e a v y ones (Fig. 3b).
24
Fig. 2. Frequency distribution of tunnel inlet size.
4. Results and discussion the f o r m u l a e d e v e l o p e d and tested by Z e n g (1983) and then s e d i m e n t discharges in the f l o w w e r e further excluded. The field m o n i t o r i n g p r o g r a m was c o n d u c t e d f r o m
4.1. Tunnel flow hydrological responses (a) Tunnel flow start time. Tunnel f l o w start time is here defined as the lag time b e t w e e n the onset o f
Table 1 Major characteristics of the tunnel systems in Yangdaogou subbasin Tunnel system
Morphometry No. of Minimum inlets tunnel length (m)
Physical characteristics Mean slope
Outlet area (m 2)
Minimum catchment area (m2)
1
4-5
2
0.35
0.13
2,410
2
1
6
1.08
0.68
2,050
8-12
132
0.57
0.13
12,700
5-9
45
0.72
0.36
3,285
3
44
0.46
0.3
3,210
254
0.36
3.10
25,600
5
6
24-25
Developed on an old landslide and the bottom of the tunnel is the contact between loess and red earth. It drains water from the gentle slopes covered with dense shrub (Caragana). Formed within loess deposits. Two layers of tunnel conduits are connecting to the outlet but the upper one is abandoned. Tunnel drains water from cultivated slopelands. Developed within the loess and the outlet is located just above the gully bottom land. Three series of inlets ranked by size up the valley slope. Tunnel catchment consists of a large proportion of terrace lands. Developed within the loess and the outlet; is also located just above the gully bottom land. Tunnel catchment underwent a great change over the monitoring period. The outlet is hanging on the cliff. Though the uppermost tunnel inlet is adjacent to the above terrace land there is no indication that overland flow enters the tunnel system. Thus the main drainage area is confmed to the vailey slope. The largest tunnel in the basin. Red earth is scattered along the gully bottom of the catchment. In the 1990 rainy season, a new inlet leading to 40 m or so tunnel conduit was formed during one single storm.
T.X. Zhu / Geomorphology 20 (1997) 255-267 700
259
(a) _.=
600
A :500 E
g
,~
f
m
E O 200
100
o
50
I
I
I
I
I
I
100
150
200
250
300
350
Day of Year
50
40
(b) ¸
[] Long Term • Mon tored Period
A
o
30
O"
u.
20
10
0-9.9
10-19.9
20-29.9
30-39.9
40-49.9
Max 30-min Intensity (ram/h)
Fig. 3. (a) Cumulative rainfall in the Yangdaogou in 1989 and 1990. (b) Frequency distribution of maximum 30-min rainfall intensity for runoff-generation storms in a long-term and the monitored period.
Table 2 Annual precipitation with various return periods Category
Wet year
Return periods Annual precipitation (mm)
200 908
After Zhang et al. (1992).
100 865
50 807
20 733
10 671
5 601
Average year
Dry year
2 483
5 399
10 334
20 299
50 240
T.X. Zhu / Geomorphology 20 (1997) 255-267
260
rainfall and the emergence of flow at the tunnel outlet. Tunnel flows occurred in twelve of the fifteen monitored storms. Overall, their response to rainfall was very fast, with the start lag times ranging from 1 to 67 min (Table 3). No distinctive difference exists between tunnel flow start time and overland flow start time, which implies that tunnel flow velocity is very fast. Although no measurements were made, we could roughly estimate it. The shortest start time for tunnel 6 was 5 min. Even if we assume that overland flow was initiated immediately and entered the last tunnel inlet after the rainfall start, the tunnel flow still had an average velocity of 0.26 m s-]. It is noted that the tunnel length of this section was actually measured by the author passing through it. (b) Tunnel flow peaks A total of 42 discharge peaks were observed from 35 complete process records at four tunnels (numbers 1, 3, 4, 6). Most tunnel hydrographs only have a single peak, which occurs within half an hour after the start of tunnel flow. The number of peaks seems to be determined by rainfall characteristics and is not affected by tunnel complexity. But exceptions did exist. For example, the storm of I 1 July 1990 was characterized by a single rainfall peak. Likewise, one discharge peak occurred in tunnels 1, 4 and 6. How-
2
,~ ~'1.5
"~'E !go,
Rainfall
I o~
J
II
I
I
I'
I
200
...;;
150 / t I t
100 u
?a O
50
i I
+
Tunnel Hydrograph
------T4
,~
0
"
60O 500 400 300 •~ 200 ._ 100 tn 0
I
J
Flume Hydrograph
L 20
40
60
80
100
120
Time (min)
Fig. 4. Rainfall, tunnel and basin outflow hydrographs of the August 11, 1990 storm.
ever, two tunnel flow peaks were produced in tunnel 3 (Fig. 4). The reason is still unclear. One possible explanation is that this is due to the runoff generation zonation. In light or medium storms, runoff in
Table 3 Tunnel and overland flow start time Date
Precipitation
Antecedent rainfall
First 10-min intensity
Runoff start time (min)
(mm)
(mm) (5 days)
(mm/min)
TI
T2
T3
T4
T5
T6
S1
F
6 August 89 10 August 89 15 August 89 16 August 89
24.8 13.1 28.7 29.1
0 24.8 13.1 28.7
0.99 0.37 0.65 0.15
10 N 21.4 N
16 N 4.2 N
4 N 18 N
8.7 N 17.5 N
13 N M N
5 N 20 N
M M M M
13 25 15 10
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
8.4 29.5 30.5 49.5 1.6 46.9 33.3 53.1
0.04 1.19 1.8 0.16 0.42 1.05 0.09 0.4
N 12 15 49 67 19 M M
N 15 8 29 40 1.0 M 22
N 13 11 59 M 9 M 30
N 12 14 12 30 11 M 25
N M 8 29 20 8 M M
N 23 17 15 10 17 18 12
110 13 12 53 73 6 M 15
M 3 11 27 12 7 6 17
11 August 90 13 August 90 28 August 90
20.2 35.4 53
3.9 24.1 0
0+63 09 1.4
20 M 13
17 29 15
23 M 23
20 13 17
M M M
20 15 15
18.3 21 21 M M 20
T = tunnel system; S = surface plot; F = experimental basin flume; M = missed events; N = no runoff occurs.
T.X. Zhu / Geomorphology 20 (1997) 255-267
the tunnel catchment is mainly generated from the cultivated slopes and the steeper valley side slopes and can be quickly directed into the tunnel system. However, in heavy storms such as the above-mentioned one, a large amount of runoff can also be generated from the e:~tensive terrace lands and takes a considerably longer time to reach the tunnel inlets than the former owing to the relatively remote locations and gentle slopes. Therefore, it formed a second discharge peak for the same rainfall peak. The uneven peaks in the flume hydrograph in Fig. 4 imply the slight difference in peak time among different tunnels. (c) Tunnel flow duration. In light and medium storms, tunnel flow duration shows only limited differences among the tunnels, and is more or less comparable to the effective rainfall duration (Table 4). Here, effective r,'dnfall is defined as the rainfall with an intensity of over 0.2 m m / m i n (Zhu et al., 1995). However, in the heavy storms, such as on 11 July 1990, 26 July 1990 and on 28 August 1990, tunnels 3 and 4 had a considerably longer flow duration than the others. This is because the runoff could be generated from the terrace lands in heavy storms and it took quite a long time to enter the
261
tunnel networks for tunnels 3 and 4, whereas the tunnel flow duration for the remaining ones was still quite similar to that in light and medium storms, owing either to relatively shorter distances between terrace land and nearest inlets or to higher slope gradients in the tunnel catchments.
4.2. Impacts of instability within tunnel systems on tunnel flow hydrology Deep-seated tunnel systems in this area are characterized by great instability. Collapses within tunnel systems are very common. Small-scale collapses may only cause oscillation of sediment concentrations in the tunnel flow and no effects on tunnel flow hydrology. However, large collapses could exert profound impacts on tunnel flow hydrology. If the collapses are extremely large or associated with surface depression or sediment deposition, tunnel systems could be totally blocked. Rapid tunnel flow could, in turn, reopen the blocked tunnel systems. Such temporal shift of tunnel systems can be detected using smoke bombs before and after storm events. In the 1989 rainy season, the outlet of tunnel 1 was connected to two series of inlets, but the major branch was blocked
Table 4 Tunnel and overland flow duration Date
Precipitation
Antecedent rainfall
Rainfall duration
Rainfall with 1 > 0.15 m m / m i n duration
Runoff duration (min)
(mm)
(mm)(5 days)
(rain)
(min)
T1
T2
T3
T4
T5
T6
S1
F
6August89 10August 89 15August89 16 A u g u ~ 89
24.8 13.1 28.7 29.1
0 24.8 13.1 28.7
45 35 170 535
45 35 60 10
25 N 21 N
29 N 41 N
58 N 37 N
32 N 45 N
31 N M N
41 N 40 N
M M M M
61 25 165
445
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
8.4 29.5 30.5 49.5 1.6 46.9 33.3 53.1
450 8 132 110 67 185 220 85
0 8 42 110 30 70 60 50
N 11 37 51 31 60 M M
N 10 9 41 14 46 M 25
N
N
7 96 71 M 103 M 26
6 51 55 28 120 M 29
N M M M M 50 M M
N 9 21 57 45 48 27 29
7 17 27 56 51 108 M 34
M 34 82 96 60 105 141 60
llAugust90 13 August90 28August90
20.2 35.3 53
3.9 24.1 0
83 914 255
30 54 140
24 M 62
21 40 42
32 M 92
28 77 83
M M M
39 52 41
36 38 42
53 M 63
T = tunnel system; S = surface plot; F = experimental basin outflow flume; M = missed events; N = no runoff occurs.
T.X. Zhu / Geomorphology 20 (1997) 255-267
262
in the 1990 rainy season. This led to a great disparity in tunnel flow discharge between 1989 and 1990 (Fig. 5). In tunnel 3, the southern branch, consisting of four tunnel inlets, was blocked throughout the 1989 and 1990 rainy seasons, but it was reopened in 1992 (DiCenzo, pers. commun., 1995). The middle branch was also blocked during the storm of 11 August 1990 and reopened by the storm of 28 Au-
60 E ~.=_ 40
1--
gust 1990, which caused the tunnel flow discharge of the former storm to be disproportionally low. In tunnel 4, three connecting tunnel inlets, about 30 m south of the tunnel outlet, were abruptly joined into tunnel 4 during the storm of 13 August 1990. Those newly joined tunnel inlets subsequently added a large amount of runoff to the tunnel system generated from slope land and terrace land.
.Pr~.tie0.~mm~
.
=
1
20
o
75 50
N
I~ N
o 100 75
Tunnel3
25
0
1°t ~ O0
Tunnel4
o
300 loo
250
Tunnel6
2OO 150
N
50
o
Date M: Missed events N: No runoff
Fig. 5. Water discharges at tunnels 1, 3, 4 and 6 in 1989 and 1990.
1
I1
263
T.X. Zhu / Geomorphology 20 (1997) 255-267
The most significant event which I observed during the two consecutive rainy seasons was the abrupt initiation of one tunnel inlet on 13 August 1990. The inlet, with a diameter of 1.5 m and depth of 1.9 m, was developed in the middle of a road, located on the upper drainage boundary of tunnel 6. Smoke bomb tests indicated that it was connected to an inlet of tunnel 6 about 40 m away! The runoff from the village and the neighbouring subbasin, which used to flow into another ba:~in via the excavated road, was redirected into tunnel 6 through the inlet and conduit of the newly developed tunnel. This led to the discharge of tunnel 6 being unusually high during the storm of 13 August 1990 (Fig. 5). After that storm, the inlet was filled in by the villagers, since it hindered traffic. As a result, discharge at tunnel 6 returned to normal in the subsequent storm of 28 August 1990. Totally blocked tunnel branches can be readily detected with smoke bombs and thereby their impacts on tunnel flow hydrology can be explicitly evaluated. However, in most cases, the tunnel systems may not be totally blocked but partially dammed or blocked first and reopened later during the same storm. In these situations, smoke bombs are useless and it is extremely dangerous to investigate by crawling into the tunnel systems after storms. Thus normally no direct evidence is available. However, in July of 1989, I did manage to pass through the last section of tunnel 6, a 76 m long tunnel conduit, and found a bridge-like constriction inside. Apparently, it used to be a dam caused by a collapse and sometime later tunnel flow penetrated the dam and formed the opening under the 'bridge'. Here, in contrast to the intended objective of this section, I use the monitored tunnel flow hydrologic processes to identify the possible partial damming during the event. Owing to the lack of direct evidence, the results presented here must be considered tentative and to be examined further in the future. In the storm of 26 July 1990 (Fig. 6), tunnel flow processes in all tunnels except tunnel 4 were characterized by an early discharge peak, which was caused by an immediate intensity peak after rainfall onset. However, discharge at tunnel 4 was very low in the first hour and peak discharge did not occur until 77 min after rainfall onset. After the peak, the discharge sharply dropped to a very low level and lasted for another 40
1.5 Rainfall
_= E
.o 0
60 I I I t I l t"
40 30
" ° "T3 -----T4 -- -- -- T6
Tunnel Hydrograph
,Oo 200| $ lso~
/
~
FlumeHydrogmph
=.;e'oto 0
I 20
40
60
80
I I 100 120 140 160 180 200
Tlrne (mln)
Fig. 6. Rainfall, tunnel and basin outflowhydrographsof the July 26, 1990 storm.
min or so. The total discharge appeared to be normal and the absence of the first peak probably resulted from partial damming, which was flushed away later by accumulated water inside the tunnel. In contrast, in tunnel 6, after the first discharge peak, tunnel flow simply stopped. It was unlikely that no runoff had been generated by the second rainfall peak from the tunnel catchment, the largest one in the basin. The total discharge from tunnel 6 during this storm was also quite low. Twelve hours later, another storm occurred, with a rainfall of 19.8 mm. Flow in tunnel 6 started 18 min after rainfall onset and the discharge was so high that the trapezoidal weir shifted. Accordingly, no water discharge data were collected, although sediment samples were still taken throughout the event. Actually, the mean rainfall intensity for the first 20 min (0.09 mm/min) was very low in this storm. Though the antecedent soil moisture was very high, it is unlikely to have produced such a high flow if no water had been trapped by damming during the previous storm. Two rainfall peaks produced four discharge peaks in tunnel 3 (Fig. 6; similar to that shown in Fig. 4), may have been caused by runoffgeneration zonation as well.
264
T.X. Zhu / Geomorphology 20 (1997) 255-267
Table 5 R 2 values in linear regression analyses between runoff discharge and effective rainfall ( l > 0.20 m m / m i n ) Overland flow
Tunnel flow
Plot No.
R 2 (including all events)
Tunnel No.
R 2 (including all events)
R 2 (discarding events affected by instability)
S-1 (zone 1) S-2 (zones l a n d 2) S-3 (zones 1, 2 a n d 3)
0.642 (30) 0.535 (34) 0.617(40)
T1 T3 T4 T6
0.075 0.691 0.257 0.125
0.622(10) 0.713 (11) 0.542 (9) 0.301 (11)
(12) (12) (11) (13)
Note: Values in parentheses represent the number of events.
To further evaluate the impact of instability on tunnel flow discharge, I compared it with overland flow discharge. Owing to the limited number of events that were monitored on the surface plots during our monitoring periods, I have used the data collected by the Shanxi Institute of Soil and Water Conservation during the period of 1963-1968 from the Yangdaogou subbasin (SISWC, 1982). Three surface plots with areas of 400, 1855 and 4167 m 2 were selected in the subbasin comprising upper slope, lower slope, and combined slope, respectively. It
was found that a good correlation exists between runoff discharge and rainfall with an intensity of more than 0.2 m m / m i n for all three surface plots (Table 5). For tunnel systems, such a good correlation could only be~found for tunnel 3. However, if we disregard thoseevents affected by tunnel instability, identified above, correlation coefficients are improved for all tunnels, especially for tunnel 1 and 4. The still poor correlation for tunnel 6 may be ascribed to unidentified tunnel instabilities within this large and complex system.
Table 6 Contribution of tunnel flow to the experimental catchment outflow Date
Rainfall (mm)
Effective rainfall (mm)
Tunnel flow a (m 3)
6 August 89 10 August 89 15 August 89 16 August 89
24.8 13.1 28.7 29.1
24.8 13.1 22 8.9
142.2 0 194.7 0
6 July 90 7 July 90 11 July 90 13 July 90 22 July 90 26A July 90 26B July 90 30 July 90
21 9.5 39.7 28.5 18.2 33.3 19.8 15.8
5.2 9.5 35.4 28.5 15.5 28.4 16.6 14.2
11 August 90 13 August 90 28 August 90
20.2 35.4 53
Total
390.1
Tunnel flow b (m 3)
Outflow (m 3)
Contributions a (%)
Contributions b (%)
224.0 0 221.8 0
280.7 36.6 330.7 57.3
50.7 0 58.9 0
79.8 0 67.1 0
0 6.1 190.0 141.1 64.3 147.5 0 80.6
0 19.3 218.2 153.6 104.5 164.6 82.5 85.8
120.5 85.6 940.6 283.5 164.5 360.4 332.8 104.5
0 7.2 20.2 49.8 39.1 40.9 0 77.1
0 22.5 23.2 54.2 63.5 45.7 25.0 82.1
18.1 22.9 44.7
20.0 353.8 189.9
23.9 486.6 219 8
79.8 389.2
25.0 48.8
29.9 56.5
307.8
1530.2
1871.6
3566.7
42.9
52.5
Notes: a Estimated from Method 1. b Estimated from Method 2.
T.X. Zhu / Geomorphology 20 (1997) 255-267
4.3. Hydrological significance of tunnel flow Tunnel flow discharge data are only available from tunnels 1, 3, 4 and 6, but they account for about 90% of the catchment area of all tunnel systems in the basin. Here, two methods are used to evaluate the contribution of tunnel flow to the basin outflow. In Method 1, only those monitored events and tunnel systems are considered which could represent the minimum contribution of tunnel flow. In Method 2, catchmen)I area and effective rainfall are used to interpolate the discharges of tunnels 2 and 5, as well as those missed events for the other four tunnels. The results are listed in Table 6. Overall, the total tunnel flow contribution to basin outflow is 42.9-52.5%, which is slightly higher than the proportion of tunnel catchments to total experimental area, 40% or so. Great interstorrn variations also exist. In the light storms with low intensities such as on 10 August 1989, 16 August 1990 and 6 July 1990, no tunnel flow was observed though overland flow was generated. Field investigations showed that overland flow generated within the tunnel catchment area x~as reduced by collapse debris or the surrounding materials of the tunnel conduits while
265
flowing into tunnel systems via inlets. Such reduction was maximized at the beginning of the rainy season and increased with tunnel length. Because we did not monitor the beginning of the 1989 rainy season, I analyze here the temporal variation of tunnel flow discharge only for the 1990 rainy season. The first runoff-generation storm occurred on July 6, with a total of 21 mm precipitation, and 120 m 3 runoff passing through the low flume, but no tunnel flow emerged from any tunnel outlets. On the following day, a total of 9.5 m m rain fell within 8 min and tunnel flow occurred in all tunnel systems except tunnel 6. Tunnel 1, with the blockage of a longer branch (70 m or so) and the opening of a shorter one (2 m or so), had an even higher runoff coefficient than the whole experimental catchment (Fig. 7). However, for tunnel 3, with an at least 45 m long conduit from the last inlet to outlet, the nmoff coefficient was still very low. No tunnel flow was observed from the outlet of tunnel 6, with a 76 m long conduit from last inlet to outlet. Even in the storm of 11 July, the third one of the rainy season, tunnel 6 contributed only 62.3 m 3 or 7.3% runoff to the experimental subbasin outflow. This was still disproportionally smaller than its catchment proportion to
25
llil T1
20
l
liT3
lilT4 •T6 • Subbas
A
15 U
E
u ~ lO e,-
6-Jul
7-Jul
11-Jul
13-Jul
Date
Fig. 7. Temporal variations in runoff coefficients of tunnels and experimental catchment at the beginning of the 1990 rainy season.
266
T.X. Zhu / Geomorphology 20 (1997) 255-267
the subbasin, 21%, whereas runoff coefficients at other tunnels were more or less comparable to the whole experimental catchment. Hamilton (1990) has also stated " f o r some tunnel outlets there was no evidence of flow during the first two events of the year [1988] whereas major discharges were directly observed from these tunnels during the third events of the rainy season." From Table 6, it can be seen that the tunnel flow contributions to the subbasin outflow are often significantly high from medium storms such as on 6 and 15 August 1989, and on 22 and 30 July 1990. This may be due to several reasons. First, after the tunnel was wetted at the beginning of the rainy season, the thick tunnel roof prevents evaporation from wetted tunnel conduits. Second, after sheet or rill flows enter the tunnel systems, infiltration losses should be reduced owing to the small contact area. Third, particle size analyses indicated that all six tunnel floors were composed of weak permeable layers with higher clay contents than the surface (Luk and Zhu, 1995), which could further reduce infiltration losses. In very heavy storms, such as those on 11, 26 and 28 July 1990, tunnel flow contributions to the outflow fall off again and are more or less comparable to the proportion of tunnel catchment areas to the basin. This is because the overland flow is generated throughout the basin and the difference in infiltration losses between tunnel systems and untunnelled surface areas is considerably reduced. Additionally, heavy storms may result in active collapses in the tunnel systems, which could trap some tunnel flow.
5. Conclusion and implications Deep-seated tunnel systems in this area act as conduits for overland flow. Soil throughflow is too shallow to reach the tunnel systems during storm events. Field experiments and investigations indicate that the wetting depth is usually less than 30 cm and rarely exceeds 50 cm during the storms (Li Gang, 1992), whereas almost all the tunnel systems are located much below this. The ground water is too deep to recharge the tunnel flow; although some tunnel inlets are 20 m deep they are still not deep enough to be close to the ground water table.
Tunnel flow does not simply mirror overland flow. First, high levels of absorption of water at the beginning of the rainy season and the high efficiency of flow delivery later create significant differences in the temporal variation of flow discharge over the season. Second, instability within the tunnel systems could make tunnel flow discharge or processes highly erratic within storms. Third, some inlets not connected to any outlets still drain water from the catchment. The trapped water is not only absorbed by surrounding materials but may also recharge the ground water, which could be a very important process in this area. Field rainfall simulation experiments conducted in this area showed that micro-pipes could also be developed close to the surface. Initiation of those shallow pipes is closely related to rill development (Bryan et al., 1978; Yair et al., 1980; Wang, 1991). However, in this area, they seem unlikely to develop into deep-seated tunnels. Instead, the huge tunnel systems are more likely to be initiated in catastrophic storms, although the subsequent storms do expand them. One tunnel inlet leading to about 40 m of tunnel conduits, was developed in a single storm on the 13th July 1990. Another example was reported from Wudu, Gansu Province in the western Loess Plateau. A total of 123 tunnel inlets were developed within an area of 0.2 km 2 during a single storm with a precipitation of 37 mm and duration of 115 min in 1962 and the volume of developed tunnel totalled around 1000 m 3 (Wang, 1989). These eroded sediments mixed with runoff and developed into so-called hyperconcentrated flow and were flushed out to stream channels via existing or newly developed conduits (to be discussed in a separate paper, Luk and Zhu, 1995).
Acknowledgements This study was funded by the Canadian International Development Agency. The author is grateful to Profs. S.H. Luk, Chen Yongzong, Cai Qiangguo and Zeng Boqing for their valuable advice during the setting up of monitoring equipment. I am also very much indebted to both Canadian and Chinese research teams for their support in the field. The manuscript significantly benefited from the com-
T.X. Zhu / Geomorphology 20 (1997) 255-267 m e n t s a n d s u g g e s t i o n s o f Prof. R. B r y a n a n d t w o a n o n y m o u s referees.
References Baillie, I.C., Fanlkner, P.H, Espin, G.D., Levett, M.J., Nicholson, B., 1986. Problems of protection against piping and surface erosion in central Tuni,;ia. Environ, Conserv. 13, 27-32. Bryan, R.B., Harvey, L.E., 1985. Observation on the geomorphologic significance of tunnel erosion in a semi-add ephemeral drainage system. Geogr. Ann. 67 (A), 257-272. Bryan, R.B., Yair, A., Hodges, W.K., 1978. Factors controlling the initiation of runoff and piping in Dinosaur Provincial Park badland, Alberta, Canada. Z. Geomorphol. Suppl. 29, 151-168. Chen, Y.Z., 1958. Tunnel formation and development in the Loess Plateau. Geography Notes, 2:25-30 (in Chinese). Cumberland, K.B., 1944. Soil Erosion in New Zealand. Soil Conservation and River Control Council, Wellington, 288 pp. Drew, D.P., 1982. Piping in the big muddy badlands, southern Saskatchewan, Canada. In: Bryan, R., Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 293-304. Fuller, M.L., 1922. Some unusual erosion features in the loess of China. Geogr. Rev. 12, 570-584. Gilman, K., Newson, M.D., 1980. Soil Pipes and Pipeflow--a Hydrological Study in Upland Wales. British Geomorphological Research Group Monogr. 1, GeoBooks, Norwich, 114 pp. Hamilton, H., 1990. Field Examination of Soil Erosion and Losses of Nitrogen and Phosphorus from the Agricultural Watershed, I. The Loess Plateau, People's Republic of China. Unpublished Ph.D. thesis, University of Toronto, 205 pp. Harvey, A., 1982. The vale of piping in the development of badlands and gully systems in southeast Spain. In: Bryan, R., Yair, A. (Eds.), Badl~nd Geomorphology and Piping. GeoBook, Norwich, pp. 317-335. Heede, B.H., 1971. Characteristics and processes of soil piping in gullies. U.S. Department of Agriculture, Forest Service Research Paper, Rm 68, 15 pp. Hughes, P.J., 1972. Slope aspect and tunnel erosion in the loess of Banks Peninsula, New Zealand. J. Hydrol. (New Zealand) 11, 94-98. Imeson, A.C., Kwaad, lz.J.P.M., Verstraten, J.M., 1982. The relationship of soil physical and chemical properties to the development of badlands in Morocco. In: Bryan, R. Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 47-70. Jones, J.A.A., 1981. The Nature of Soil Piping--a Review of Research. British Geomorphological Research Group Monogr. 3, GeoBooks, Norwich, 301 pp. Jones, J.A.A., 1982. Experimental studies of pipe hydrology. In:
267
Bryan, R. Yair, A. (Eds.), Badland Geomorphology and Piping. GeoBook, Norwich, pp. 355-370. Jones, J.A.A., 1987. The effects of soil piping on contributing areas and erosion patterns. Earth Surf. Process. Landf. 12, 229-248. Jones, J.A.A., Crane, F.G., 1984. Pipeflow and pipe erosion in the Maesnant experimental catchment. In: Burt, T.P., Wailing, D.E. (Eds.), Catchment Experiments in Fluvial Geomorphology. GeoBooks, Norwich, pp. 55-72. Li Gang, 1992. Experimental Study on Rainfall Infiltration--one of the Fundamentals of Hillslope Amelioration in Loess Plateau. Unpublished Ph.D. thesis, Institute of Geography, Chinese Academy of Sciences (in Chinese with English abstract). Liu, D.S., 1964. Loess Deposits in the Middle Reaches of Yellow River. Science Press, Beijing, 234 pp. (in Chinese). Luk, S.H., Zhu, T.X., 1995. Tunnel erosion and sediment production in the hilly loess region, North China. (in preparation). McCaig, M., 1983. Contributions to storm quickflow in a small headwater catchment: the role of natural pipes and soil macropores. Earth Surf. Process. Landf. 8, 239-252. Morgan, R.P.C., 1972. Observation on factors affecting the behavior of a first-order stream. Trans. Inst. Br. Geogr. 56, 171-185. SISWC (Shanxi Institute of Soil and Water Conservation), 1982. Rainfall, runoff and sediment data (1955-1981). Unpublished (in Chinese). Thorp, J., 1936. Geography of the soils of China. National Geological Survey of China, Nanking. Waish, R.P.D., Howells, K.A., 1988. Soil pipes and their role in runoff generation and chemical denudation in a humid tropical catchment in Dominica. Earth Surf. Process. Landf. 13, 9-17. Wang, B.K., 1989. Factors causing tunnel erosion. Acta of Soil and Water Conservation, 3(3): 84-90 (in Chinese). Wang, G.P., 1991. Modeling Soil Erosion from Hillslope in the Hilly Loess Region, China. Unpublished M.Sc. thesis, University of Toronto, 117 pp. Yair, A., Bryan, R.B., Lavee, H., Adar, E., 1980. Runoff and erosion processes and rates in the Zin Valley badlands, Northern Negev, Israel. Earth Surf. Process. 5, 205-225. Zeng, B.Q., 1983. Methods for soil and water losses monitoring in the small catchment. Soil and Water Conservation of Shanxi, 2:11-19 (in Chinese). Zhang, Z.G., Zen, B.Q., Jia, Z.J., 1992. Precipitation characteristics in Wangjiagou basin, western Shanxi. In: Wang, F.T. (Editor), Soil Erosion Management and Application of Geographical Information Systems in the Loess Plateau region, western Shanxi. pp. 1-12 (in Chinese). Zhu, T.X., Cai, Q.G., Zeng, B.Q., 1995. Runoff generation in a semi-arid agricultural catchment: field and experimental studies. Submitted to Journal of Hydrology (under review).