Revegetation can alter catchment water balance and result in soil desiccation. Large-scale revegetation took place in the Loess Plateau of China to control soil erosion and improve environmental conditions. However, the dynamic nature of soil moisture in response to revegetation under different climatic conditions is still unclear mainly due to lack of long-term in situ observations. To overcome this challenge, a biophysically based ecohydrological model (WAVES) was used to examine the effects of revegetation on soil moisture. Results showed that trees consume more water (100% of precipitation) than shrub (97.6%) and grass (98.3%), and therefore are more likely to result in soil desiccation. No runoff occurred under the tree scenario, while for shrub and grass, runoff accounted for 2.4% and 1.7% of precipitation, respectively. In areas with mean annual precipitation (MAP) less than 400 mm, tree planting resulted in soil water deficit, while in areas with MAP exceeding 600 mm, no soil water deficit occurred. Within this MAP range (400 < MAP < 600 mm), this could lead to soil water deficit during dry years. Extending this analysis to the entire Loess Plateau, 40% of the region will face reduced soil moisture when converting cropland to trees.
INTRODUCTION
China experienced severe droughts in 1997 and serious floods in 1998, causing serious economic and environmental damage. In the wake of these disasters, the Chinese government took unprecedented conservation measures (Xu & Cao 2001), one of which was the Grain for Green Program (GGP, also known as the Conversion of Cropland to Forest and Grassland Program) introduced in 1999 to protect the degraded environment (Zhang et al. 1999). The objective of this program was to convert cropland to plantations and grassland on steep slopes by compensating farmers with subsidies. The ‘Grain for Green’ Program aimed to achieve a target of converting 146,700 km2 of cropland to plantations by 2010 and total planned investment exceeded US$40 billion (SFA 2003). Although the project was designed to address environmental issues (e.g. climate change, droughts, water crisis) in China, it could set up a good example for other nations to achieve both socioeconomic and environmental sustainability through revegetation strategies. The success of the ‘Grain for Green’ Program has important global implications (Liu & Diamond 2005).
The focus of this research was the Loess Plateau, which is the largest highly erodible area on earth (Peng 2001), where the pilot phases of the GGP were implemented. Many studies have reported positive effects of GGP on the Loess Plateau. For example, Xin et al. (2008) and Cao et al. (2009) concluded that vegetation cover on the Loess Plateau exhibited an overall ascending tendency during the period 1998–2005. Feng et al. (2010) documented that recovery of vegetation was effective in controlling soil erosion. However, recent studies have also raised concerns about the possibility of water scarcity and soil desiccation following revegetation in water limited areas and this can affect the sustainability of further development of GGP (Wang et al. 2011a, 2011b). Feng et al. (2012) detected that water yield in 38% of the Loess Plateau has decreased as a result of land cover change alone. Deep soil layers in the Loess Plateau are regarded as ‘soil reservoir’ with high water-holding capacity, which could sufficiently mitigate drought and supply the necessary water for plant growth. This function has been threatened by afforestation due to increased evapotranspiration (Chen et al. 2008b). Based on in situ measurements, studies showed clearly that afforestation can lead to soil desiccation (Li et al. 2008; Wang et al. 2011a, 2011b; Zhang et al. 2012). For example, Yang (2001) discussed the formation of dry soil layer and indicated that soil desiccation is the negative effect of afforestation. Cao et al. (2009) indicated that soil water content in an afforestation plot at depths of 0 to 1.0 m and 1.0 to 2.0 m, were 63.2% and 42.8% lower than in the abandoned plots, respectively. Wang et al. (2009) showed that plantations had depleted soil moisture below 2 m, as a result they rely on present year precipitation for transpiration. Li et al. (2008) reported that average soil moisture in a 0–10 m soil profile of the forestland was lower than that of nearby grassland. Fan et al. (2010) conducted a field study of the soil water balance under different land use patterns and discovered that trees may reduce soil moisture, but after trees were removed, soil water was replenished. Yang (2001) detected a response of soil moisture to land use and afforestation at a depth of 2 m, and showed that the deep soil moisture decreased more than 35% after land use conversion, and a soil moisture deficit appeared, and high planting density was found to be the main reason for the severe soil moisture deficit. However, there are few studies about long-term inter-annual changes in soil moisture under different vegetation types due to the lack of long-term in situ observation and simulation, the dynamic mechanism of water consumption of vegetation related to soil desiccation is still unknown. Quantitative analysis is needed for further research especially on a basis of the long term relationship of water supply and demand (Xu 2005; Jung et al. 2010). Sankaran et al. (2005) reported that trees are unlikely to form closed-canopy woodland with less than 650 mm of precipitation, and afforestation could result in soil moisture deficit especially in water limited semi-arid regions, however it is still not clear how revegetation will affect soil moisture under different precipitation gradients in the Loess Plateau.
In this study, an ecohydrological model (WAVES) was used to evaluate the effect of revegetation on soil moisture dynamics in the Loess Plateau. One advantage of the WAVES model (Zhang et al. 1996) is that it properly represents the soil moisture dynamics by dynamically linking hydrological processes with vegetation growth at the spatial scale so that it can accurately simulate development of LAI, canopy transpiration, rooting dynamics and soil water stress on both transpiration and growth at the plot or catchment scale (Cheng et al. 2014). The WAVES model has been successfully tested against field observation of soil water content, groundwater, and evapotranspiration especially in Australia and China (Salama et al. 1999; Wang et al. 2001; Crosbie et al. 2010; McCallum et al. 2010; Cheng et al. 2014). On the Loess Plateau of China, the model also showed good performance in evaluating water use efficiency, and soil water content (Kang et al. 2003).
The objective of this study was twofold: (1) to evaluate the effects of revegetation on soil moisture on the Loess Plateau; (2) to investigate whether revegetation in given precipitation conditions could result in long-term soil moisture reduction.
MATERIAL AND METHODS
Study area
The parent material of the soil is loess with a depth of 50–200 m depending on topography. The loess in this region has a porosity of about 50%, the gravimetric field capacity and wilting percentage of soil water are 20–24%, and 3–6%, respectively (Yang & Shao 2000). Average soil water content in the topsoil (0–100 cm) ranges from 7% to 13% and details of the study catchment can be found in Wang et al. (2013).
The catchment experienced major land use change. Before 1999, cropland accounted for 29.6% of the total catchment area. After implementation of the GGP, the cropland on steep slopes were gradually abandoned and converted to trees, shrubs, and grass with only 0.11% of the cropland remaining (Liu et al. 2012).
Field measurements and data collection
Four typical land cover types with similar slope positions (upper position), aspects (west), and slope degrees (25°) were considered. Soil samples were collected for particle size and bulk density measurements. The plant cover and height were also measured. The planted tree was R. pseudoacacia, the shrub was S. pubescens, the grass site was dominated by beard Andropogon, and the main crop was corn (Zea mays).
Soil moisture and weather data were measured with six S-SMC-M005 and six S-SMC-M006 soil moisture and temperature sensors fixed on H21 Soil Moisture and Temp Logger Systems. The soil moisture sensors were installed at six depths: 10, 20, 40, 60, 80 and 100 cm below the ground surface. The soil moisture sensor is capable of measuring volumetric saturation values of 0–100%, with an accuracy of ±1.0% and a resolution of 0.1%. Soil moisture data were collected by a HOBO weather station logger every 10 min, while other meteorological variables (air temperature, relative humidity, and wind velocity at a height of 2 m above the ground) were recorded every 30 min.
Model application
WAVES (Dawes & Hatton 1993; Zhang et al. 1996) is a one-dimensional process-based model that simulates the process of water, energy, and solute movement in soil–plant–atmosphere continuum on a daily time step. It integrates soil, canopy-atmosphere with a consistent level of process detail, thus the model is well suited to investigations of hydrological and ecological responses to changes in land management and climate variation. It is particularly used in soil water analysis by Yang et al. (2002) and Kang et al. (2003). Detailed descriptions of which can be found in Zhang & Dawes (1998).
WAVES requires three data sets as inputs: meteorological data, soil parameters, and vegetation parameters.
Meteorological data
Soil and vegetation parameters
Soil water movement is simulated with the Richards equation (Richards 1931), and the BW soil model (Broadbridge & White 1988) was used to generate values of soil water potential, water content, and hydraulic conductivity to allow WAVES model to simulate the movement of soil water with the Richards equation. Table 1 shows the BW soil parameters used in this research, which includes hydraulic conductivity Ks (m/d), saturated water content θs (m3/m3), water content at the wilting point θr (cm3/cm3), capillary length scale λc (m), and characteristic soil water retention curve, C. Some soil physical parameters were determined from available databases on the basis of soil texture at the field sites. Soil hydraulic conductivity, water content at the wilting point and saturated water content were obtained from Kang et al. (2003). In this research, the soil was divided into three layers of different thickness for all vegetation species, the dominant soil type of each layer was quite different (Table 1). The vegetation parameters were obtained from the WAVES user manual (Dawes et al. 1998). We listed the optimal parameters of different vegetation types, when vegetation conversions occur, we will fix the soil parameters, and then change the vegetation parameters according to Table 2.
Layer . | Texture . | Depth (cm) . | Ks (m/d) . | θs (m3/m3) . | θr (cm3/cm3) . | λc (m) . | C . |
---|---|---|---|---|---|---|---|
1 | Silt loam | 0–40 | 0.10 | 0.20 | 0.07 | 0.25 | 1.20 |
2 | Sandy loam | 40–80 | 0.30 | 0.21 | 0.08 | 0.30 | 1.10 |
3 | Loam | 80–300 | 0.80 | 0.21 | 0.11 | 0.30 | 1.05 |
Layer . | Texture . | Depth (cm) . | Ks (m/d) . | θs (m3/m3) . | θr (cm3/cm3) . | λc (m) . | C . |
---|---|---|---|---|---|---|---|
1 | Silt loam | 0–40 | 0.10 | 0.20 | 0.07 | 0.25 | 1.20 |
2 | Sandy loam | 40–80 | 0.30 | 0.21 | 0.08 | 0.30 | 1.10 |
3 | Loam | 80–300 | 0.80 | 0.21 | 0.11 | 0.30 | 1.05 |
Ks is the saturated hydraulic conductivity; θs and θr are the saturated and air-dry volumetric water contents, respectively; λc is a capillary length scale related to sorptivity; and C is a shape parameter for the soil moisture characteristic.
. | . | . | . | Vegetation types . | |||
---|---|---|---|---|---|---|---|
No. . | Parameter . | Unit . | Source references . | Tree . | Grass . | Shrub . | Crop . |
1 | 1 minus albedo of the canopy | – | Brutsaert (1982) | 0.85 | 0.85 | 0.85 | 0.80 |
2 | 1 minus albedo of the soil | – | Brutsaert (1982) | 0.85 | 0.85 | 0.85 | 0.80 |
3 | Rainfall interception coefficient | m d−1 LAI−1 | Vertessy et al. (1996) | 0.001 | 0.002 | 0.001 | 0.001 |
4 | Light extinction coefficient | – | Monteith & Unsworth (2008) | −0.45 | −0.85 | −0.45 | −0.70 |
5 | Maximum carbon simulation rate | kg C−2 d−1 | Collatz et al. (1991) | 0.2 | 0.035 | 0.04 | 0.04 |
6 | Slope parameter for the conductance model | – | Leuning (1995) | 0.9 | 0.9 | 0.9 | 1.0 |
7 | Maximum plant available soil water potential | m | Hillel (1980) | −200 | −150 | −200 | −300 |
8 | IRM weighting of water | – | Wu et al. (1994) | 2.1 | 2.0 | 2.1 | 1.5 |
9 | IRM weighting of nutrients | – | Wu et al. (1994) | 0.5 | 0.5 | 0.5 | 0.5 |
10 | Ratio of stomatal to mesophyll conductance | – | – | 0.2 | 0.8 | 0.8 | 0.8 |
11 | Temperature when the growth is 1/2 of optimum | °C | Farquhar et al. (1980) | 15 | 20 | 20 | 20 |
12 | Temperature when the growth is optimum | °C | Farquhar et al. (1980) | 25 | 30 | 25 | 25 |
13 | Year day of germination | D | – | 100 | 100 | 100 | 100 |
14 | Degree-daylight hours for growth | °Chr | Charles-Edwards (1982) | 30,000 | 30,000 | 30,000 | 30,000 |
15 | Saturation light intensity | μmoles m−2 d−1 | Wu et al. (1994) | 1,200 | 1,800 | 1,800 | 1,800 |
16 | Maximum rooting depth | M | Hatton et al. (1992) | 1.0 | 1.0 | 1.0 | 1.0 |
17 | Specific leaf area | LAI kg C−1 | Charles-Edwards (1982) | 50 | 24 | 40 | 30 |
18 | Leaf respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.001 | 0.0002 | 0.0001 | 0.0015 |
19 | Stem respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.0006 | −1 | −1 | −1 |
20 | Root respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.0001 | 0.0001 | 0.0001 | 0.0002 |
21 | Leaf mortality rate | fraction of C d−1 | Running & Coughlan (1988) | 0.0005 | 0.005 | 0.0065 | 0.00045 |
22 | Above-ground partitioning factor | – | Running & Gower (1991) | 0.3 | 0.4 | 0.6 | 0.6 |
23 | Salt sensitivity factor | – | – | 1.0 | 1.0 | 1.0 | 1.0 |
24 | Aerodynamic resistance | s d−1 | Brutsaert (1982) | 15 | 30 | 20 | 20 |
25 | Crop harvest index | – | – | 0.001 | 0 | 0 | 0 |
26 | Crop harvest factor | – | – | 0.001 | 0 | 0 | 0 |
. | . | . | . | Vegetation types . | |||
---|---|---|---|---|---|---|---|
No. . | Parameter . | Unit . | Source references . | Tree . | Grass . | Shrub . | Crop . |
1 | 1 minus albedo of the canopy | – | Brutsaert (1982) | 0.85 | 0.85 | 0.85 | 0.80 |
2 | 1 minus albedo of the soil | – | Brutsaert (1982) | 0.85 | 0.85 | 0.85 | 0.80 |
3 | Rainfall interception coefficient | m d−1 LAI−1 | Vertessy et al. (1996) | 0.001 | 0.002 | 0.001 | 0.001 |
4 | Light extinction coefficient | – | Monteith & Unsworth (2008) | −0.45 | −0.85 | −0.45 | −0.70 |
5 | Maximum carbon simulation rate | kg C−2 d−1 | Collatz et al. (1991) | 0.2 | 0.035 | 0.04 | 0.04 |
6 | Slope parameter for the conductance model | – | Leuning (1995) | 0.9 | 0.9 | 0.9 | 1.0 |
7 | Maximum plant available soil water potential | m | Hillel (1980) | −200 | −150 | −200 | −300 |
8 | IRM weighting of water | – | Wu et al. (1994) | 2.1 | 2.0 | 2.1 | 1.5 |
9 | IRM weighting of nutrients | – | Wu et al. (1994) | 0.5 | 0.5 | 0.5 | 0.5 |
10 | Ratio of stomatal to mesophyll conductance | – | – | 0.2 | 0.8 | 0.8 | 0.8 |
11 | Temperature when the growth is 1/2 of optimum | °C | Farquhar et al. (1980) | 15 | 20 | 20 | 20 |
12 | Temperature when the growth is optimum | °C | Farquhar et al. (1980) | 25 | 30 | 25 | 25 |
13 | Year day of germination | D | – | 100 | 100 | 100 | 100 |
14 | Degree-daylight hours for growth | °Chr | Charles-Edwards (1982) | 30,000 | 30,000 | 30,000 | 30,000 |
15 | Saturation light intensity | μmoles m−2 d−1 | Wu et al. (1994) | 1,200 | 1,800 | 1,800 | 1,800 |
16 | Maximum rooting depth | M | Hatton et al. (1992) | 1.0 | 1.0 | 1.0 | 1.0 |
17 | Specific leaf area | LAI kg C−1 | Charles-Edwards (1982) | 50 | 24 | 40 | 30 |
18 | Leaf respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.001 | 0.0002 | 0.0001 | 0.0015 |
19 | Stem respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.0006 | −1 | −1 | −1 |
20 | Root respiration coefficient | kg C kg C−1 | Running & Coughlan (1988) | 0.0001 | 0.0001 | 0.0001 | 0.0002 |
21 | Leaf mortality rate | fraction of C d−1 | Running & Coughlan (1988) | 0.0005 | 0.005 | 0.0065 | 0.00045 |
22 | Above-ground partitioning factor | – | Running & Gower (1991) | 0.3 | 0.4 | 0.6 | 0.6 |
23 | Salt sensitivity factor | – | – | 1.0 | 1.0 | 1.0 | 1.0 |
24 | Aerodynamic resistance | s d−1 | Brutsaert (1982) | 15 | 30 | 20 | 20 |
25 | Crop harvest index | – | – | 0.001 | 0 | 0 | 0 |
26 | Crop harvest factor | – | – | 0.001 | 0 | 0 | 0 |
Model calibration
Scenario design
Effects of vegetation on soil moisture were evaluated by considering different vegetation change scenarios: cropland to shrub, cropland to grass, and cropland to trees. In the simulations, daily meteorological data in the period of 1980 to 2010 were used. Changes in average soil moisture over 3 m soil layer under three scenarios were compared at seasonal and annual scales.
RESULTS
Calibration of the WAVES model
Changes in seasonal water balance under different vegetation scenarios
Besides, seasonal changes in interception and evapotranspiration in the four vegetation covers are apparent and they responded positively to precipitation (Figure 4). When vegetation germination began, monthly interception and transpiration were generally low (0.05 mm–2.4 mm and 0.45 mm–2.6 mm, respectively). Most of interception and transpiration occurred during the growing season (i.e. 77–92% of interception and 73–85% of transpiration).
Average annual water balance under different vegetation scenarios
Soil moisture response under different precipitation regime
DISCUSSION
Implication for upscaling
Comparing with other models for investigating the processes of soil water dynamics, the WAVES model strikes a good balance between complexity and accuracy of prediction in water, energy and carbon processes. This is the first study to evaluate changes in long-term soil water storage under different vegetation scenarios on the Loess Plateau of China.
The successful application of the WAVES model to the Yangjuangou catchment suggests that the model is capable of representing hydrological processes on the Loess Plateau. According to the methodology presented for assessing the average changes in groundwater recharge under a future climate in the Murray-Darling Basin, Australia (Crosbie et al. 2010), the output of the WAVES modelling in our research can be used to produce regression equations between average rainfall, average annual evapotranspiration, and soil moisture for each combination of soil type, vegetation, and climate. Using the set of regression equations and the set of annual average rainfall rasters for each climate scenario, a series of annual average evapotranspiration, and soil moisture rasters could be developed to the whole Loess Plateau for historical and different future climate scenarios. Therefore this research could be considered for upscaling to the whole Loess Plateau region, and then the effects of revegetation on soil moisture at regional scale can be examined.
Effects of revegetation on water budget
Miralles et al. (2010) detected global canopy interception from satellite observations, and found that interception is responsible for the evaporation of approximately 13% of the total incoming rainfall over broadleaf evergreen forests, 19% in broadleaf deciduous forests, and 22% in needle leaf forests. In arid and semi-arid region where most rainfall occurs in low frequency, interception by canopy can amount to more than 10% of the annual rainfall. In this research, the simulated interception of shrub, grass, crop, and trees is approximately 14.2%, 15.9%, 12.7%, and 16.2%, respectively. Our results are consistent with previous studies, and fells within the scope (10–23%) of previous calculations for different vegetation interception losses (Wang et al. 2011a, 2011b). Clearly rainfall interception is a significant component of the water budget and quantifying the magnitude of rainfall interception is essential if we are to understand the future impact of afforestation on the water budget.
On average the simulated ratio of evapotranspiration to precipitation was 83.4%, 82.4%, 86.3%, and 83.8% for shrub, grass, crop, and trees, respectively. Evapotranspiration loss fell in the scope (62–87%) for the corresponding vegetation type on the Loess Plateau (Wei et al. 1998), although our results showed a little higher evapotranspiration for trees compared with other studies showing evapotranspiration loss of 76% (Wang et al. 2014). However, both studies indicated that interception and evapotranspiration were the major pathway of water loss, accounting for over 97% of the precipitation for shrub and grass. Therefore interception and evapotranspiration are important in water flux and have the major impact on soil moisture (Good & Caylor 2011). Aligned with Bellot et al. (1999) and Kyushik et al. (2005) conclusion that evapotranspiration is a main cause of depletion of soil moisture, our research also found that evapotranspiration and interception accounted for 100% of precipitation during cropland abandonment to trees for the last 30 years, which indicates that all water from precipitation evaporated, and the soil water storage was depleted, soil water deficit occurred in this scenario, and no runoff was produced. While for shrub and grass, runoff accounted for 2.4%, and 1.7% of precipitation, respectively. The runoff ratio produced by our WAVES-based simulation was relatively comparable with previous research in the same catchment (Liu et al. 2012), which reported a ratio of about 1.55%, and 1.57%. Excluding errors produced in the process of field observation and model simulation, the difference of runoff may be attributed to vegetation morphology and structure, which can change the characteristics of rainfall that reaches the soil surface (Lorens & Domingo 2007).
The characteristics of water loss from the soil profile were also detected in previous studies at Changwu (annual precipitation of 584 mm) and Wuqi County (annual precipitation of 451 mm) of the Loess Plateau (Li 2002). Furthermore, Chen et al. (2008a) showed that the soil moisture deficit could become serious in dry and normal years. Our study demonstrated that it is necessary to maintain infiltration or reduce evapotranspiration by, for example, selecting appropriate vegetation types in order to avoid soil water deficit.
Implication for ‘Grain for Green’ program
Revegetation such as afforestation on the Loess Plateau resulted in an increase in vegetation cover (Xin et al. 2008), an improvement in soil nutrient levels (Wang et al. 2011a, 2011b), and in controlling soil erosion (Feng et al. 2012). Implementation of afforestation at large scale throughout the arid and semi-arid region, ignoring differences in climate and hydrology, in the end reduced soil moisture, therefore soil water availability was not sufficient to sustain forests in many areas (Wang et al. 2011a, 2011b; Zheng et al. 2014), ultimately affecting forest survival (Cao et al. 2011). Hydrological response to vegetation restoration varied across the Loess Plateau and, as it has a strong north-south gradient in precipitation and terrain, afforestation remains challenging (Feng et al. 2012). There are a number of factors that can affect afforestation and the impact on soil moisture and these include precipitation (Wang et al. 2009; Liu & Shao 2015), planting density and productivity (Chen et al. 2008a, 2008b; Yang et al. 2012). Therefore different vegetation conditions (i.e. structure, density, growth age, and species of vegetation), water balance pattern, ecosystem-carrying capacity must be carefully considered and should be given more attention (Sankaran et al. 2005).
CONCLUSIONS
We investigated the effects of vegetation characteristics on soil moisture reduction under different precipitation conditions on the Loess Plateau in North Western China. In order to quantify their long-term effects on soil water dynamics, the WAVES model was applied for the 1980–2010 period, and three revegetation scenarios were considered. The following conclusions can be drawn from this study.
The simulated soil moisture at the daily scale showed good agreement with observed values with relative error of only 1%, indicating that the WAVES model generally exhibits a good performance.
As simulated interception ratios of shrub, grass, crop, and trees are about 14.2%, 15.9%, 12.7%, and 16.2%, respectively, clearly rainfall interception is a significant component of the water budget. Quantifying the magnitude of rainfall interception is essential if we are to understand future impact of afforestation on the water budget.
Trees consume more water (100% of precipitation) than grass (97.6%) and shrubs (98.3%), and therefore no runoff is generated, and thus more likely to result in soil desiccation.
Areas with MAP less than 400 mm afforestation could result in serious soil water deficit; in areas with the MAP exceeding 600 mm, no soil water deficit would occur; within this MAP range (400 < MAP < 600 mm), tree planting could lead to soil water deficit during dry years. About 40% of the Loess Plateau has MAP of less than 400 mm where plantation development will face reduced soil moisture. About 10% of Loess Plateau with MAP more than 600 mm showed soil water surplus and soil water seems sufficient to sustain tree growth. The other 50% of Loess Plateau (400 < MAP < 600 mm) will face reduced soil moisture in dry years under the tree planting scenario.
ACKNOWLEDGEMENTS
We acknowledge, with gratitude, the financial support from National Natural Science Foundation of China (41390464); The CAS-CSIRO Cooperative Research Program (GJHZ1502); the International Postdoctoral Exchange Fellowship Program (20130002); and the China Postdoctoral Science Foundation (2012M520403).