Soil Organic Carbon、Water Variability And Plant Diversity In Desert Landscape In The Middle Reaches Of Heihe River Basin

The ecological environment in arid desert regions is extremely frail. As the plant growthmatrix, soil quality directly affects the stability of desert ecosystems. Soil organic carbon(SOC) has an important influence on the physical, chemical, and biological properties of soil.Its concentration levels determine the growth of ground surface vegetation, and play animportant role in the global carbon cycle. Soil moisture in desert environments acts as the keylimiting factor in vegetation development. The spatial and temporal variability of soilmoisture limits the types and amounts of vegetations growth, and largely determines theorganization and function of desert ecosystems.Based on a large number of in-situ soil samples collections and laboratory analysis, andthe use of classical statistics and geostatistical methods, this dissertation studied the spatialvariability of SOC and illustrated its distribution patterns, and explored the current level andstocks of SOC across this Gobi desert landscape in the middle reaches of Heihe River Basin.Through the long-term monitoring of soil water content in the field, we investigated thespatiotemporal variability and temporal stability characteristics of surface soil moisture, andidentified their controlling factors, and also studied the scaling of temporal stability forsurface soil moisture. In addition, by analyzing the spatial patterns of plant communityspecies diversity in this area, we discussed the influence of soil properties on the speciesdiversity. The main results are listed as follows:(1) Overall, SOC concentrations in Gobi desert were low, and SOC was mainly storedin0-30cm layer. SOC density to depths of20and40cm for this40km2area was estimated at0.42and0.68kg C m-2, respectively. With the increase of soil depth, variability of SOCshowed a decreasing trend. Variability of SOC increased as the sampling area expanded andcould be well parameterized as a power function of the sampling area. SOC concentrationvariability in the surface layer was more sensitive to the expansion of area. The relativelycoarse fractions, i.e. sand, silt, and stone contents, had the largest effects on SOC variability.Experimental semivariograms of SOC were best fitted by exponential models. It indicated astrong spatial dependence for SOC concentrations at all depths in the study area. Thecomposition of the parental material and the weathering that led to its formation may be responsible for the strong spatial dependence of SOC.(2) The surface-layer soil water content in this desert area was tightly linked to rainfallevents. The standard deviation of soil water content decreased logarithmically as mean watercontent decreased, and the coefficient of variation of the mean water content exhibited aconvex upward pattern. Soil wetness is a significant factor influencing the spatial variabilityof soil moisture for finer-textured soils. The CV of soil water content also increased with thesize of area and could be well parameterized as a power function of the size. Thesemivariograms of the soil water contents were optimally fitted by exponential models for alldates, showing strongly to moderately spatially dependent. The spatial dependence of soilwater content changed over time, with stronger patterns of spatial organization in drier andwetter conditions of soil wetness and stochastic patterns in moderate soil water conditions.The dominant factors regulating the variability of soil water content changed fromcombinations of soil and topographical properties (bulk density (BD), clay content andrelative elevation) in wet conditions to combinations of soil and vegetational properties (BD,clay content and shrub coverage) in dry conditions.(3) Surface-layer soil water content showed strong temporal stabilities at both the gridscale and the transect scale. A single representative sampling location (RSL) can be identifiedto accurately estimate the field-or transect-mean soil water content. Among the methods forestimating the field mean moisture content using the representative locations, the minimalstandard deviation of relative difference (SDRD) applied with a constant offset provided thebest results. A minimal SDRD played a more important role than a value of mean relativedifference (MRD) close to zero. Soil properties, i.e., BD, total porosity (TP), soil-particlecomposition, and SOC were the main factors influencing the temporal stability of soilmoisture in the desert region. Locations with relatively higher sand contents tended to havemore pronounced temporal stabilities. Among all kinds of soil textures in this region, loamysand displayed the most temporally stable characteristics, and sandy clay loam displayed theleast temporally stable characteristics. Soil water content was more temporally stable in dryperiods than in wet periods. The dynamics of soil moisture during wet periods should be paidmuch more attentions in the preliminary simulation scheme that designed for the RSLsidentification. Sampling frequency can be properly reduced for the coarse-textured soil.(5) The spatial variability and temporal stability characteristics of surface-soil watercontent were scale-dependent. Soil moisture temporal stability of overall spatial pattern andthe individual locations increased with increasing sampling spacing; while with the increasingof sampling extent, the temporal stability of overall spatial pattern increased and the temporalstability of individual locations decreased. Sampling extent had a stronger effect on the temporal stability of surface-layer soil moisture than sampling spacing. However, the specificpatterns of scaling differed among different parameters. For example, the standard deviationand coefficient of varaibility of soil water content did not significantly change with samplingspacing (p>0.05), but they increased significantly with increasing sampling extent and couldbe well parameterized as a power function of the sampling extent (p <0.01). With increasingsampling spacing or sampling extent, the mean values of the Spearman rank correlationcoefficient both showed increasing trends, but for the formed a exponential equation couldexpress the relationship (p=0.06) and for the latter a logarithmic equation could express wellthe relationship (p <0.01). For the range of MRD, it increased by a power function with bothincreasing sampling spacing and decreasing sampling extent (p <0.01), while for SDRD, itdecreased exponentially with the increase in sampling spacing (p <0.01) and decreased by apower function with the decrease in sampling extent (p <0.01).(6) In desert regions, the methodology of using soil moisture at a temporally stablelocation in one area (forecasting area) to predict mean soil moisture in another adjacent ordistant area (target area) is available. The variability of different properties has minorinfluence on the prediction of soil water contents between different areas, and a slightnon-linearity of the ratio of mean soil moisture between forecasting area and targert area canbe ignored in soil moisture upscaling. The mean absolute relative error (MARE) decreased bya power function with the increase of early measurement times in the target area. Theprediction was much more accurate when larger areas were involved for mean soil moistureprediction. The MARE for prediction from larger areas to smaller areas is less than thatobtained from smaller areas to larger areas. For the same area of forecasting areas, MAREdecreased with increasing area of the target areas.(7) There is totally42species of plant belonging to16families39genera in this desertregion. The most abundant families were Composite, Chenopodiaceae, Zygophyllaceae,Leguminosae, and Poaceae. The Patrick richness index (R), Shannon-Wiener diversity index(H), Simpson dominance index (D), and Pielou evenness index (J) all showed good spatialpatterns. The R and H were accorded with gaussian models, and the D and J were accordedwith spherical models. During the formation of plant community structure, the variabilities ofD, H, and J were determined together by the structural factors and the random factors, and thevariability of J was mainly determined by the structural factors. Species richness and diversityin this desert region were chiefly influenced by the soil physical properties, i.e., soil watercontent, BD, soil-particle composition, saturated hydraulic conductivity, and TP, and were notinfluenced by SOC (p>0.05). There was no association between species evenness and anysoil properties (p>0.05). Based on a large number of field samplings and experimental measurements, a series ofissues on soil organic carbon, water content variability and plant diversity in a Gobi desertregion were explored. The reliable spatial data update the SOC storage database for the Gobideserts, and can support an important aspect in understanding the role of Gobi desert soils inthe global carbon cycle. The findings add to the knowledge about the spatio-temporalvariability and temporal stability of soil water content in Gobi deserts, which have importantimplications for the setting of sampling locations, optimization of sampling scheme,and transformation of different spatial scales in the hydrological process simulations. Therelevant results are also of great value guidance in the vegetation protections andeco-environment management of Gobi deserts.