The formation of a dried soil layer (DSL), on the Loess Plateau of China, is an integrated response of water cycle under the circumstance of regional climate, soils, topography, hydrological processes. It mainly results from the excessive depletion of deep soil water by non-indigenous or natural vegetation through excessive evapotranspiration combined with long-term insufficient amounts of rainfall. The occurrence of DSLs potentially limits the development and sustainability of the ecological environment on the Loess Plateau.
Aiming at the ecological problems caused by soil water shortage on the Loess Plateau Region, especially the soil desiccation phenomenon in the soil-vegetation-atmosphere transfer system, the main objectives of this dissertation were: (1) to explore the spatial distribution characteristics of soil basic physical parameters on the Loess Plateau; (2) to monitor the formation and development processes of DSLs on the Plateau; (3) to investigate the spatial variability and patterns of DSLs as well as the dominate factors of DSLs across the entire Plateau; and (4) to illustrate the distribution characteristics of deep soil water (21 m) in the different zones of the Loess Plateau. Based on the research approach of intensive sampling design → field survey → field sampling and measurement → laboratory analysis, first, we monitored the dynamic evolution processes of DSLs; then, we pre-selected 382 sampling sites across the Loess Plateau (~ 620 000 km2) based on mapped information using a sampling grid of 40 km × 40 km in 2008, and in the next year (2009), we selected 266 sampling sites in the eight zones of the Plateau. For all the sites, soil water samples were collected by using a soil auger (5 cm in diameter), in 10 cm increments down the soil profile to a sampling depth of > 5 m (the deepest depth was 21 m); moreover, undisturbed soil cores from the soil surface (0-5 cm) and subsurface layer (20-25 cm), and disturbed soil samples from the 0-20 cm and 20-40 cm depths were collected to determine soil properties potentially related to DSLs; in addition, the environmental conditions of the sampling sites were recorded in details, i.e., land use type, soil type, vegetation species and coverage, vegetation age, crop yields, and groundwater level, etc.
After determined scientifically the evaluation indice of DSLs (represented by three indices: DSL thickness, DSLT; DSL forming depth, DSLFD; and①② mean SWC ③within the DSL, DSL-SWC), we conducted the research by using classical statistics and geostatistical methods with related software (i.e., Arcgis9.2, GS+ 7.0, SPSS13.0, SigmaPlot2001, Genstat12.1, etc). The main results are listed as follows:
(1) Soil basic physical properties (including soil particle composition, field capacity, permanent wilting point, available water capacity, bulk density, saturated soil water content, capillary soil water content, saturated hydraulic conductivity, deep soil water contents at different soil depth below 200 cm depth), at surface and subsurface layers, demonstrated a distinct spatial variation characteristics and distribution patterns both in horizontal and vertical directions, across the Loess Plateau. All parameters showed a moderate spatial dependence (except saturated hydraulic conductivity, and sand content, field capacity, capillary soil water content at upper layer, were strongly dependent). The semivarigrams of soil particle composition and field capacity can be best fitted with Gaussian model, and the best fitted model of saturated soil water content and the deep soil water content at 200-800 cm depths was spherical model, while for saturated hydraulic conductivity was exponential model. The values of range for all parameters varied from 75 km (surface saturated hydraulic conductivity) and 684 km (subsurface clay content). The spatial distribution patterns of soil hydraulic parameters on the Loess Plateau result from the long-term interactive influence of regional external conditions, interior elements, and human activities.
The dynamic development process of DSLs and its desiccation extent were significant influenced by land use, vegetation types, and plant growth age, in the wind-water erosion crisscross region on the Loess Plateau. The extent of soil desiccation for different land use types followed the sequence of artificial Caragana korshinskii shrub land > artificial alfalfa land > natural grass land > farm land. The degree of soil desiccation under natural vegetation was generally less than that under non-indigenous plant species, and thus, the use of natural vegetation succession management principles would possibly reduce soil desiccation during vegetative restoration. Forming rate of DSLs thickness was dependent on vegetation type: DSLs formed after two years of alfalfa (Medicago sativa) growth and three years of Caragana korshinskii growth; after four years of growth, DSLs under alfalfa were thicker than those under C. korshinskii, but after 31 years the DSL thickness under C. korshinskii (440 cm) exceeded that formed under alfalfa (300 cm).
more persistent DSLs occurred below a 100 cm thick upper soil layer that was seasonally dried and replenished by rainfall. Densities of root length, weight, and surface area, and the average root diameter of soybean (Glycine max), alfalfa, Stipa bubgeana, and C. korshinskii all decreased with increases in soil depths below 20 cm. Correlations between soil water content (SWC) and root indices, and various soil physical and chemical properties, were generally weaker within the DSL layers than within the whole soil profile. The only significant correlation was between soil organic carbon and SWC under alfalfa (r = 0.627, P < 0.05). Soil desiccation may thus interfere with these typical inter-relationships occurring within the whole soil profile.
(3) The dried soil layers, occurred on the Loess Plateau, distributed widely and demonstrated an obvious spatial variation and special distribution patterns. There was strong spatial variation (CV = 110%) in DSLs, which had a mean thickness of 160 cm occurring at a mean soil depth of 270 cm. Land use had a significant impact on both DSLT and DSLFD (P < 0.001).
statistic analysis showed that DSLT indicated strong spatial dependence while DSLFD had moderate spatial dependence. The values of range and nugget ratio for DSLT and DSLFD were 33.9 km and 125 km, 79% and 50%, respectively. The DSL was generally thicker (> 170 cm) in the western Loess Plateau region and in a central area (170 m to 220 cm) between Shaanxi and Shanxi boundary. Where irrigation was used along parts of the Yellow River (i.e., Ningxia and Neimeng irrigation districts) and near rivers in the interior (i.e., Fenhe irrigation district, Guanzhong plain), DSLT was considerably thinner or non-existent due to the higher water inputs. At regional scales, the dominate factors of DSLT were land use, rainfall, soil type and slope gradient, while for DSLFD were only land use, rainfall, and soil type.
(4) In different zones of the Loess Plateau, the profile distribution of soil water content at 0-21 m depth, and the relationships between soil water content and soil particle composition, soil organic carbon were different. At 0-600 cm depth, soil water content generally showed a trend of decreasing-increasing; while in the whole 21 m profile, the vertical trend of soil water content was waving change, however, it can be divided into different layers to describe the profile soil water characteristic based on the vegetation type and profile distribution of plant roots.
can determine the lower bound of DSL by means of measuring deep soil water content (21 m in our study). Our investigation in seven typical zones of the Plateau showed that the maximum of low bound for DSL was 1325 cm (Wuqi County, Shaanxi province; forest land), while the minimum was 450 cm (Suide County, Shaanxi province; farm land).
For the 21 m soil profile in Shenmu (shrub land), Suide (farm land), and Wuqi (forest land), soil water content correlated significantly with soil organic carbon, clay, silt, and sand content (r > 0.552, P = 0.01); while in Guyuan (grass land), the correlation was not significant. Soil organic carbon significantly correlated with clay, silt, and sand content in Shenmu and Wuqi sites (P < 0.01), and it only correlated negatively with silt content in Guyuan site (P < 0.05). In Suide site, the correlations between soil organic carbon and clay, silt, and sand content were not significant.
(5) In the different climatic regions of the Loess Plateau—arid (P < 250 mm), semiarid (250 mm < P < 500 mm), and semihumid (500 mm < P < 800 mm), land use and plant characteristic had a significant impact on DSLs. (a) In the arid region, land use had no significant effect on DSLs but there were significant effects between farmland and grassland or forests (P < 0.05) in the semiarid region. In the semihumid region, DSLs under forests had a significantly greater DSLT and SFC than those under farmland and grassland (P < 0.05). Therefore, optimizing land use can mediate DSL formation and development in the semiarid and semihumid regions of the Loess Plateau and in similar regions elsewhere. (b) The development of DSLs under trees and grasses was generally more severe in the semiarid region than in the semihumid region. In each climatic region, the extent of DSLs depended on the plant species (e.g., native or exotic, tree or grass) and growth ages. The spatial pattern of DSL-SWC in the different climatic regions was an integrated result of large-scale and small-scale factors and their interactions.
On the Loess Plateau, a dried soil layer under forest land has been generally formed (102 of 125 sampling sites), and the degree of soil desiccation was serious (DSLFD = 140 cm，DSLT = 304 cm，DSL-SWC = 7.92% < FC = 10.21%). Based on the potential 28 correlated variables of DSL, we developed the regression models for the three indices of DSL (DSLT, DSLFD, DSL-SWC), combining the methods of correlation analysis, principal component analysis, minimum data set, and multiple regression. The models had a high precision, especially for DSL-SWC (Adjusted R2 = 73%).
ld capacity, bulk density, clay content, slope gradient, and aridity degree impacted DSL significantly, which can be used to predict the three indices of DSL at a certain confidence level. On the Loess Plateau, using these five variables to predict DSLs under forest land can improve the study efficiency.
(7) Small-scale factors (i.e., vegetation type, growth age, altitude, slope aspect and position) had a great impact on DSLs when the large-scale factors were uniform. (a) According to the study conducted in the interior district (Wanrong County, Shanxi province): under the circumstance of topography, vegetation characteristics, and management measure were uniform, DSLs indicated a “layer effect”—a desiccated soil layer with similar thickness in horizontal direction. (b) Based on the study conducted in the hilly and gully region (Ansai County, Shaanxi province), vegetation types and growth age had highly impact on DSLs. DSLT and DSL-SWC differed significantly under different vegetation types (soybean, Subgen. Artemisia, C. korshinskii, and Robinia pseudoacacia Linn). With the increasing of growth age for C. korshinskii, and Robinia pseudoacacia Linn, the profile change of soil water content existed an inflexion (26 and 15 years, respectively)—before the inflexion, soil water content and soil desiccation extent increased with the increasing of growth age; while after the inflexion, soil water content slightly increased and soil desiccation extent gradually alleviated, although the soil layer still belonged to the DSL. Ascertaining the inflexion of plants is very important to vegetation construction, water management, and soil water restoration. (c) The study conducted on Suide County, Shaanxi province, showed that slope aspect and position influenced DSL greatly—DSL was more severe in shady slope than in sunny slope; while in the same slope aspect, DSL was more severe in upslope than in down-slope and valley bottoms. (d) DSLFD showed an altitude gradient according to the study conducted on the Luochuan County, Shaanxi province.
he basis of understanding the spatial distribution characteristics of soil basic physical parameters and the dynamic development of DSL, utilizing knowledge of the spatial variation characteristics of DSLs, spatial distribution patterns of DSLs, dominant factors affecting DSLs at the regional scale, and regression models of DSLs is benefit to the plant species selection, overall arrangement, and management measures under different climate, soils, and terrain on the Loess Plateau. This information also is helpful to keep the balance between rainfall (soil water input) and evaportranspiration (soil water output), and then enable scientifically based policies (e.g., sustainable land use management, distribution and/or nature of regional revegetation projects) to be made that would alleviate the process of soil desiccation and sustain development of the economy and restoration of the natural environment. Moreover, these results can also be useful to the modeling of the regional water cycle and related eco-hydrological processes. This study may have importance both in theory and in practices, for example, DSL avoiding and reclaim, soil erosion controlling, vegetation restoration and eco-environment reconstruction.
Key Words：The Loess Plateau; Soil desiccation; Spatial variability; Deep soil moisture; Land use; Vegetation construction; REML; Statistical model