Catchment-scale Available Soil Moisture Spatial-temporal Variability In The Hilly Areas Of The Loess Plateau

Rainwater utilization is an effective means for relieving soil-water loss and watershortage on the Loess Plateau. Available soil moisture (ASM) is one of the majorcomponents of rainwater potential for the Loess Plateau, and it is also critical to theecological restoration and agriculture of this region. In fact, the ASM is highly variable inspace and time due to the influences of meteorology, soils, topography and vegetation.However, the spatial-temporal features of ASM are not fully understood. In order to meetthe need for efficient use of rainwater in the hilly areas of the Loess Plateau, thisdissertation focused on the spatial-temporal variability of ASM in a catchment namedYuanzegou catchment located in Northern Shaanxi province, based on soil moisturedatasets from2009to2012and other datasets relating to soils, topography and vegetation.By coupling methods of classical statistics, geostatistics and modeling approaches, weinvestigated:(1) the spatial-temporal variability of ASM across scales (hillslope, gully andcatchment) in the catchment;(2) the temporal stability characteristics of ASM at variousscales;(3) the spatial structure of ASM and its seasonal features at catchment scale;(4) thequantitative relations of ASM between hillslopes and gullies;(5) the modeling of thespatial-temporal variability of ASM. The main results were listed as follows:(1) Apparent seasonal and inter-annual features were observed for ASM and soilmoisture (SM) in hillslopes and gullies. During dry seasons, ASM and SM showedrelatively low values, while they showed relatively high values in wet seasons. Fordifferent years, ASM and SM at subsurface layers decayed gradually from2009to2012.The ASM showed different spatial variation features with SM. The coefficient of variancefor ASM was almost two times of that for SM although similar standard deviations wereobserved for them. The standard deviation for ASM and SM increased first and then decreased with the increase of mean water contents. The mean water content at whichstandard deviation peaked was~20%for SM and~10%for ASM. However, therelationship between mean water contents and coefficient of variance was different for SMand ASM. It is worth noting that the micro-topography including ridges, plane surfaces andpipes significantly affect the spatial distribution of ASM and SM, and pipes showed thesignificantly (p<0.05) high SM values as compare to ridges and plane surfaces.(2) A new metric for identifying time stability location, named RMSE wasintroduced. Time stability of ASM and SM for hillslopes and gullies were analyzed byusing the new metric and others. The results showed that both ASM and SM showedconsiderable time stability, whereas the time stability features for gullies and hillslopeswere different. For sampling points at hillslopes, the time stability degree of ASM waspositively correlated with that of SM, however, this was not observed at gullies. The timestability of ASM for hillslopes behaved differently at various scales. At the land use scale,ASM and SM showed very similar time stability features while SM showed stronger timestability than ASM at catchment hillslope scale. The micro-topography also significantlyaffected the time stability of ASM and SM, and ridges indicated significantly (p<0.05)higher time stability than pipes and plane surfaces.(3) We introduced the concept of extend time stability analysis, which denotesestimating spatial mean soil moisture contents for a study site through soil moisture valuesof one single sampling location away from the study site. Then we quantitatively analyzedthe relationship of ASM between hillslopes and gullies through extended time stabilityanalysis, observation operators and random combination analysis. In particular, three linearmethods (LRG、MRD and LRS) and one nonlinear method (CDF matching) were used fordefining observation operators. Overall, extended time stability analysis and observationoperators showed low estimation errors than random combination method. Nevertheless,extended time stability analysis and observation operators are applicable only whenprevious ASM datasets are available. For different observation operators, linear operatorsshowed better temporal transferability while nonlinear method had better estimationaccuracy. However, when no previous datasets are available, random combination methodcould estimate spatial means with certain accuracy, and found that there is a limited gain inestimation accuracy when more than10upland locations are randomly selected. (4) The semivariance of ASM of the Yuanzegou catchment could be well fitted bythe spherical model. The semivariance parameter showed obvious seasonal feature in termsof nugget, sill, range, and spatial heterogeneity ratio. For nugget, the decreasing order inlight of magnitude for various seasons is summer, spring and autumn; for range, thedecreasing order is summer, autumn and spring; and for spatial heterogeneity ratio, it isspring, autumn and summer. This indicated that spatial variability of ASM was relativelylow in summer but the spatial correlations were strong. The mapping of ASM showed thatordinary kriging method could relatively well characterize the spatial structure of ASM atthe catchment scale. Generally, gullies showed higher ASM than hillslopes, and forhillslopes, north-face slopes had higher ASM than south-face slopes. The spatial structurealso differed seasonally. In autumn, ASM showed apparently different spatial structure ascompare to summer and spring.(5) We developed two models, i.e., precipitation-ASM model and soil water balancemodel, for ASM modeling. The results showed that these two models could reproducerelatively well the temporal evolutions of ASM with similar estimation accuracies.Considering the precipitation-ASM model is needs less inputs, we recommended thismodel for ASM modeling in our study site. Based on precipitation-ASM model, wecalculated the ASM storage at different precipitation frequency for gullies, hillslopes andcatchment, with the initial ASM storage in2010as initial input. The results showed thatgullies had the lowest ASM storage independent of precipitation frequency with values of102.4–126.6mm, while hillslopes had the highest values, from113.9mm-138.1mm.These analyses improved the understanding of ASM spatial-temporal variations,spatial structure of ASM, the quantitative relations of ASM between gullies and hillslopes,and the modeling of ASM. The results of this dissertation could provide insights into thecalculation and evaluation of rainwater harvesting potential, the efficient use of rainwater,and vegetation reconstruction on the Loess Plateau.