S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137 131
Fig. 3. Relationships between the vegetation coverage VC, and albedo AD, , and between the aboveground biomass
W
d
, dry weight, g m
− 2
and albedo. The vegetation coverage was eye-estimated with the quadrat method. The aboveground
biomass is monthly averaged value obtained by the quadrat har- vest method. Letters A, B, C, D, E, and F, represent the un-
grazed plot, the lightly grazed plot, the moderately grazed plot, the overgrazed plot, the unfenced grassland, and the mobile sand
dune, respectively. The solid lines shown are the lines of best fit with Ad=33.11 exp−0.0067VC n=25, r
2
= 0.90 and Ad=28.35
exp−0.0022W
d
n=23, r
2
= 0.57.
evident in June when the experiment started and the canopy heights were lower, and it was very obvious
in August when sustained grazing had occurred and vegetation was in well-grown status.
Linear relationships between mean wind speeds at the highest measurement point, u
5.0
m s
− 1
and fric- tion velocity, u
∗
m s
− 1
in daytime over the grazing experiment plots and the mobile dune are shown in
Fig. 6. Friction velocity linearly increased with in- crease of wind speed. For a given value of u
5.0
, friction velocity decreased with increasing grazing intensity.
As a result of changes of canopy structure, e.g., de- crease of vegetation cover and canopy height, caused
by grazing, slope of the regression equations decreased with increase of grazing intensity. This relationship
can also be expressed as the drag coefficient, which was defined as C
d
= u
∗
u
5.0 2
. C
d
decreased with in- crease of grazing intensity as shown in Table 2. For
example, C
d
was 0.0137 at Plot A 4 August 1994, 0.0062 at Plot B 9 August 1994, 0.0038 at Plot C
23 August 1994, and 0.0034 at Plot D 11 August 1994, respectively. C
d
of the mobile sand dune was lower than those of the grazing plots.
Also shown in Table 2 are the surface roughness lengths, z
o
of the measured sites, which decreased with increasing grazing intensity. For example, z
o
was 12.86 cm at Plot A 4 August 1994, 2.27 cm at Plot B 9 August 1994, 0.53 cm at Plot C 23 Au-
gust 1994, and 0.32 cm at Plot D 11 August 1994, respectively.
4. Discussion and conclusions
4.1. Albedo and grassland desertification Reflectivity albedo, defined as the reflected frac-
tion of incident solar of natural surfaces, depends on many factors such as amount of cloud, sun an-
gle, and surface features Monteith and Unsworth, 1990. The surface albedo clearly depends on veg-
etation cover, and vegetation growth status amount of aboveground biomass as shown in Fig. 3. The
fact that albedo Ad decreases with canopy height H
c
Ad=29.28×exp–0.0091H
c
, n=24, r
2
= 0.54
as shown in Table 2 may also be used to explain the dependence of albedo upon vegetation condition.
Surface soil moisture may affect albedo, especially when rains wet soil. Because of bareness at the dune
and at some places of the overgrazed grazing plot, rain-wetted surface soil at these sites will be soon de-
pleted when exposing to strong sunshine and winds. On the contrary, the same rain-wetted soil at the sites
with larger cover of vegetation can be retained rel- atively longer due to protection of vegetation even
132 S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137
Table 3 Daytime integrated values MJ m
− 2
per day of energy components, fraction of the net radiation versus the solar radiation, partitioning of the net available radiation into the sensible heat and latent heat fluxes, and Bowen ratios BOW at the measured sites
a
Plots Dates
R
s
R
n
G H
lE R
n
R
s
GR
n
R
n
-GR
s
HR
n
-G lER
n
-G BOW
A 10 June 1994
29.0 15.2
2.1 8.2
4.9 52
14 45
63 37
1.67 1 July 1993
29.4 18.1
2.1 61
12 54
4 August 1994 19.9
12.7 1.7
6.5 4.5
64 13
55 59
41 1.44
8 August 1993 26.3
16.6 1.4
5.1 10.1
62 8
58 34
66 0.50
22 August 1992 24.2
15.6 2.5
64 16
54 B
8 June 1994 22.2
11.1 2.7
50 24
38 5 August 1993
18.2 12.8
2.0 5.9
3.0 70
16 59
55 45
1.97 9 August 1994
17.3 10.8
1.6 6.1
3.1 62
15 53
66 34
1.97 10 August 1992
20.3 12.5
3.6 62
29 44
C 4 June 1994
26.5 12.5
2.6 7.9
2.0 47
21 37
80 20
3.95 8 August 1992
17.4 9.5
1.9 55
20 44
23 August 1994 21.5
12.4 4.7
3.1 4.6
58 38
36 40
60 0.67
D 2 June 1994
25.3 10.1
4.6 4.7
0.8 40
46 22
85 15
5.88 30 July 1993
24.2 17.3
3.3 10.6
3.4 71
19 58
76 24
3.12 6 August 1992
21.5 12.0
2.7 7.5
1.8 56
23 43
81 19
4.17 11 August 1994
15.7 7.6
4.5 1.1
2.0 48
59 20
35 65
0.55 E
18 May 1991 28.2
14.0 4.5
9.3 0.2
50 32
34 98
2 46.50
21 June 1991 29.0
17.2 4.1
9.4 3.7
59 24
45 72
28 2.54
2 August 1991 26.5
16.3 5.0
4.1 7.2
62 31
43 36
64 0.57
13 September 1991 21.6
11.3 3.0
52 27
38 F
12 June 1994 24.8
9.2 4.3
4.6 0.3
37 47
20 94
6 15.33
16 July 1991 25.9
11.1 4.4
5.1 1.6
43 40
26 76
24 3.19
2 August 1994 21.1
8.6 3.1
5.3 0.2
41 36
26 96
4 26.50
18 August 1992 22.5
9.2 4.1
41 45
23
a
The definitions of Plots A, B, C, D, E, and F are the same as those in Table 2. The empty spaces mean that data were not available because of instrument failure.
though the vegetation consumes much soil water. This feature determines how long albedo will return
to its pre-rain level after a substantial rain. Therefore, grazing, overgrazing in particular, causes signifi-
cant change of grassland vegetation status and hence changes albedo. Our grazing experiment demonstrates
that the albedo increases with an increasing grazing intensity. Enclosure enhances the restoration of vege-
tation Zhao et al., 1997, and hence makes the albedo decrease.
From the grazing experiment we find that deserti- fication did not occur at the lightly grazed grassland
Plot B. Desertification appeared partially at the mod- erately grazed plot Plot C during summer and of-
ten occurred during the winter period, when plants were dormant, rain events few, and strong winds oc-
curred often. At the overgrazed plot Plot D, vege- tation cover remained lower during all four seasons
of the year due to the intensive browsing and tram- pling by sheep, here wind erosion was very severe, and
nearly 0.5 ha of experimental grassland had been de- sertified by the end of September 1994. According to
Zhu’s four classes of desertification Zhu et al., 1989, Plot C was moderately desertified, at least in winter
and spring, while Plot D was severely desertified, and even very severely desertified in some places. Hence,
with respect to vegetation cover and stock carrying ca- pacity of the grassland, albedo of the overgrazed plot
Plot D has already reached or exceeded a threshold beyond which desertification may occur. The experi-
ment demonstrates that this critical value of albedo is around 30 for the experimental grassland.
In addition, we should recognize that grassland de- sertification is seasonal, i.e., desertification progresses
S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137 133
Fig. 4. Changes in albedo and the net radiation to solar radiation ratio R
n
R
s
of the lightly grazed plot Plot B and the overgrazed plot Plot D before and after rain in August 1994. There were
62.0 mm thunder shower in the afternoon of August 6, and 45.9 mm from the late afternoon of August 13 to the early morning of
August 14, indicated by R.
more slowly in growing seasons than in non-growing seasons dormant periods when few rainfall and
frequent heavy winds create favorable conditions leading to desertification. Therefore, at the moder-
ately grazed plot desertification did not occur in the growing season, but occurred to some extent in the
dormant periods. Additionally, since the grassland is sandy, sustained overgrazing will ultimately turn the
grassland into mobile sand dunes an extreme form of desertification where albedo is ca. 34–38. In other
words, during the initiation and progression of grass- land desertification, the albedo varies from 30 to 38.
4.2. Heat fluxes and grazing intensity With each increment of grazing intensity, plant
height and vegetation cover decrease, and albedo in- creases. The increase of albedo of the grassland will
cause a decrease in the net radiation that participates in the turbulent exchange of heat and mass within the
near surface boundary layer. Ratio of net radiation or net available energy to solar radiation decreases with
increase of grazing intensity.
Pattern of the partition of net available energy into sensible heat flux and latent heat flux in grassland is
apparently affected by the grazing intensity although it is subject to other environmental factors such as
Fig. 5. Normalized wind profiles in daytime above the surfaces of the measured sites in June upper panel and August lower panel
of 1994. X-axis is mean wind speeds normalized by the wind speed at the highest measurement point. Y-axis is the logarithmic heights
of measurement height minus zero plane displacement height m. Horizontal bars represent ±1 s.e. Plots A, B, C, D, and F, represent
the ungrazed plot, the lightly grazed plot, the moderately grazed plot, the overgrazed plot, and the mobile sand dune, respectively.
Only wind speeds more than 1 m s
− 1
were included.
soil moisture and plant phenology, which profoundly affect net radiation through altering albedo. Evapora-
tive fraction in the net available energy has a tendency to decrease with increasing intensity of grazing due
to reduction of vegetation cover. Correspondingly, sensible heat fraction tends to increase. Therefore,
increase of evaporative fraction will indicate either a densely vegetated surface or a surface with higher
134 S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137
Fig. 6. Relationships between mean wind speeds at the highest measurement point and friction velocity in daytime over the grazing
experiment plots and the mobile dune. The definitions of Plots A, B, C, D, E, and F are the same as those in Fig. 5. Only wind
speeds more than 1 m s
− 1
were included. The lines of best fit shown in the upper panel are y=0.087x−0.0072, r
2
= 0.85, n=25
for Plot A on 10 June 1994, y=0.074x+0.025, r
2
= 0.95, n=24 for
Plot B on 8 June 1994, y=0.044x+0.034, r
2
= 0.68, n=22 for Plot
D on 2 June 1994, and y=0.048x−0.016, r
2
= 0.94, n=25 for Plot
F on 12 June 1994, respectively. The lines of best fit shown in the lower panel are y = 0.103x+0.070, r
2
= 0.92, n=26 for Plot A
on 4 August 1994, y=0.083x−0.014, r
2
= 0.95, n=22 for Plot B
on 9 August 1994, y=0.061x+0.0024, r
2
= 0.96, n=16 for Plot C
on 23 August 1994, y=0.057x+0.0042, r
2
= 0.97, n=24 for Plot
D on 11 August 1994, and y=0.053x+0.0029, r
2
= 0.99, n=23 for
Plot F on 2 August 1994, respectively.
moisture. Increase of sensible heat flux is attributable to stronger winds or higher surface temperatures,
which occur often at the overgrazed plot and the mobile sand dune plot.
Fraction of soil heat flux in net radiation may increase with increasing grazing intensity. Violent
diurnal fluctuations of surface soil temperature at the overgrazed plot and the mobile sand dune plot are
associated with more gain of heat from the net radia- tion. Rain-fed surface moisture at the overgrazed plot
and the mobile sand dune plot may lose more rapidly mainly by evaporation due to higher surface soil tem-
peratures. On the contrary, at densely vegetated plots, rain-wetted surfaces may lose water slowly mainly
through transpiration due to mulching of vegetation. Rapid lose of surface moisture may also increase
probability of wind-erosion because wind-erosion is closely related with soil moisture see Hu, 1991.
4.3. Vegetation status, wind regime and initiation of grassland desertification
Under increasingly intense grazing, in addition to marked decrease of grass layer height and coverage,
the vegetation and propagatative growth of the palat- able grasses is restricted due to selective grazing by
sheep Nemoto et al., 1992, 1994; Zhao et al., 1997. Field investigation also shows that some perennials
such as Pennisetum centrasiaticum and Cleistogenes squarrosa in the overgrazed plot were replaced by
some annuals such as Setaria viridis and Aristida adscensionis. When intense grazing and trampling ex-
ceed the ability of resistance and resilience of grasses, then the vegetation becomes degraded. Degradation
of vegetation is one of the most pronounced indi- cators of grassland desertification Grainger, 1992;
Bullock et al., 1994. In addition, when entered the third year experiment stage, the overgrazed plot was
invaded in some places by Agriophyllum squarrosum, which dominates in the mobile dunes and is also one
of indicating species for desertification.
The heavy trampling by sheep not only affects the normal growth of plants but also renders the soil
surfaces compact. Hardness values, measured by a hardness meter in our study, were 2.9, 3.5, 5.5, and
8.6 kg m–
2
for the ungrazed Plot A, lightly grazed Plot B, moderately grazed Plot C, and overgrazed
Plot D plots, respectively. As the surfaces become harder, infiltration rate of rain decreases, and runoff
forms and accumulates in depressions after rain- storms, which increase the risk of water-erosion and
S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137 135
wind erosion in the bare places, and further affect the heat and water budgets above the community.
In order to efficiently and reasonably utilize the grassland while simultaneously restricting the inten-
sity of usage to less than the capacity of grassland to resist overgrazing and also preventing desertification,
therefore, definition of the proper grazing intensity or stock carrying capacity is essential. We propose that
this safe or critical stocking rate be about three to four sheep or sheep equivalents per ha for the grassland
studied, based on the principles that the grassland can produce sufficient biomass to meet the requirement of
grazing sheep, and that the grazing intensity is prop- erly regulated to avoid desertification at the same time
Zhao et al., 1997.
Grazing, especially overgrazing, can affect the wind regime over the grazing experimental plots through al-
tering the surface roughness length. As shown in Fig. 6, larger wind speeds occur near the surface of the
mobile dune than the grassland, and wind speeds near the surfaces increase with increasing grazing intensity
due largely to reduction of canopy height and vegeta- tion cover by sheep that leads to decrease of surface
roughness lengths. At bare soil surfaces, e.g., at the dune and at some places of the overgrazed plot, larger
wind will play a very important role in initiation of aeolian process.
Both the drag coefficient and the surface rough- ness length can be used to express the effectiveness of
the canopy to absorb downward momentum Thom, 1975. It is clearly shown in our grazing experiment
that this effectiveness increases with decrease of graz- ing intensity. As a result of absorption of momentum
by vegetation, wind speeds are lower over the surfaces of the grazing plots than over the dune surface. The ex-
tent of reduction in wind speed will be dependent upon the grazing intensity. Downward momentum recipient
on the surface of grassland will enhance the vertical mixing near the surface while the momentum incident
upon the surfaces of the dune or desertified grassland encourage aeolian process. In other word, vegetation
layer protect the surfaces against direct wind action that may give rise to wind erosion, while removal of
vegetation by grazing may render it possible.
Wind erosion is a process in which wind acting on the surfaces deflates, entrains and abrades sand
grains, and is affected by the friction velocity Bag- nold, 1941; Anderson and Haff, 1988. Among other
forces, lift and shear stress are most important in ae- olian process Bagnold, 1941; Chepil and Woodruff,
1963. A higher vertical wind-speed gradient near the surface leads to lifting sand grains as a result
of the Bernoulli effect, and the shear stress to move the sand grain forward Pye and Tsoar, 1990; McE-
wan and Willetts, 1993. Wind erosion can occur only when wind speeds arrive in or exceed a certain
threshold to detach sand grain. Hu’s wind tunnel simulation experiment 1991 demonstrates that this
threshold is ca. 5 m s
− 1
for the sand in studied area. Threshold, however, is not the unique factor to trig-
ger wind erosion. He also suggested that in order to initiate wind erosion the requirements that vegetation
cover be less than a critical value of ca. 60 and soil moisture be less than a critical value of ca. 2
be met.
Ground surfaces became bare and exposed to wind and water due to overgrazing and trampling by sheep,
so that they were subjected to wind erosion under strong winds and to water erosion under rainstorms.
In the ungrazed, lightly grazed and moderately grazed plots, wind shear cannot act directly on ground sur-
faces and hence erosion is avoided due to the oc- currence of vegetation and its effective absorption of
downward momentum. In contrast, at the mobile dune and overgrazed plot wind erosion occurs readily be-
cause there often exists a 3–5 cm dry sand layer with water moisture less than 2 and lower vegetation
cover.
In the wind erosion process, wind-eroded patches appear first. The patches originate from the bare
places resulting from selective grazing of more palat- able plants by sheep. The wind-eroded patches expand
into a larger sheet or depression, meanwhile, eroded sand is moved to and accumulates in surrounding
places, hence leading to more severe desertification of grassland. Patch-like occurring of desertification
indicates the heterogeneity of its progression. Addi- tionally, note that although trampling by sheep is not
favorable to wind erosion due to hardening of the surfaces, because of both the sandy substrate features
of the grassland and the severe destruction of vegeta- tion by stock grazing, the land surface is persistently
subjected to wind erosion. Furthermore, compaction is one of symptoms of soil degradation, and therefore
can also be regarded as one indicator of grassland desertification Grainger, 1992.
136 S.G. Li et al. Agricultural and Forest Meteorology 102 2000 125–137
5. Summary
