| The temporospatial variations of the hydrological variables (HVs) in the hydrological system which is a complex dynamic system is an important topic in the hydrological research. Studies on this topic can help to improve our understanding of the critical hydrological processes of water cycle in a watershed, and our comprehending of their effects on local water and ecological environment. Most of the previous studies were focused on individual HVs, e.g., precipitation, river runoff, soil moisture or groundwater level, etc., or on combination of a few HVs, e.g., precipitation and streamflow, or precipitation and groundwater level etc. None of previous studies has involved all of the main HVs and the relationship among the temporal variations of different HVs. One may ask how the pressure head and water flux are auto-correlated and how their correlation envolves when precipitation infiltrates into the subsurface, seeps down in the unsaturated zone, flows in the saturated zone, and discharges as baseflow to a surface water body. In addition, one needs to know what kind of processes the infiltration at the land surface, the recharge at the water table, and the baseflow to a river are in order to better manage and assess the water resources in a watershed.In this thesis we try to answer some of these questions. First, based on numerical solutions of the moment equations and Monte Carlo (MC) simulations, the temporospatial variations of the water flow in a two dimensional unsaturated-saturated system (USS) cross section were studied. The time series, variance, auto-covariance/auto-correlation function, and power spectrum of the pressure head (Ψ(t)) and/or specific flux (q(t)) were obtained and analyzed. The USS receives infiltration from precipitation through the top boundary, and water flows downward in the unsaturated zone to the saturated zone in which groundwater flows laterally and discharges out of the system through the lower left constant total head boundary. The results show that fluctuations of Ψ(t) and q(f) are temporospatially nonstationary in the homogeneous system under white noise infiltation. The temporal nonstationarity is due to the deterministic initial conditions while the spatial nonstationarity is because of the damping effect and boundary conditions of the USS. The correlation of Ψ/(t) or q(t) changes spatially and temporally:the later in the process and the deeper in the system, the better the correlation. It is also found that the USS is a low pass filter which damps the fluctuations of Ψ(t) or q(f). The heterogeneity enhances the fluctuations of Ψ(t), controls the fluctuations of Ψ(t), and reduces its short-term correlation.Secondly, the effects of different temporally stationary and fractal infiltrations, different saturated hydraulic conductivities and different initial conditions on the temporospatial variations of Ψ(t) in the USS were studied. The correlation length of the stationary infiltration (λ1) does not change the basic composition stages of the stationarity of Ψ(t) and can affect only the length of each stage, mainly the length of the nonstationary stage; the magnitude of λ1 does not affect the basic form of the damping effect of the USS and changes only its strength; λ1 does not radically change the correlation structure of Ψ(f) and change only the lengths and correlation of different frequency sections of the power spectrum. The correlation strength of fractal infiltration (β) affects the stationarity of Ψ(t). the duration of the nonstationary stage of Ψ(t) increases withβ and finally the Ψ(t) becomes a nonstationary process;β affects the form of the damping effect of USS;β affects the correlation structure of Ψ(f) and as β increases, Ψ(f) gradually becomes a temporal fractal. The saturated hydraulic conductivity (Ks) does not affect the basic composition stages of stationarity of Ψ(t) and change only the length of each stage; Ks does not affect the basic form of the damping effect of USS and change only its strength; Ks changes the correlation structure of y/(t) significantly but not thoroughly. The duration of nonstationary stage affected by the randomness of initial condition is limited, and the randomness of initial condition does not affect the correlation structure of Ψ(t).Finally, based on an integrated model of surface water-groundwater flow in the Sagehen Creek watershed, the temporospatial variations of precipitation (P), infiltration (I), actual evapotranspiration (ET), recharge (R), baseflow (BF), streamflow (SF), groundwater level (GL), and soil moisture (SM) of the watershed were studied with their time series and power spectra. It is found that the fluctuations of P, I, ET, R, BF, and SF at the watershed scale and the fluctuations of P, I, ET, and R at the hydrological response unit (HRU) scale are all temporal scaling. The fluctuations of GL are scaling too. But the spectra of ET, R, BF and GL have breaks at low frequencies. The temporal variations of SM are scaling at high frequencies and tend to be white noise at low frequencies, and their spectra are of breakpoints at about /= 0.05/day. The breaks of the spectra of ET, R, BF and SM are due to the non-scaling damping effect of the unsaturated zone on the fluctuations at low frequencies while the breaks of the spectra of GL are mainly caused by the boundary effect at the stream. The hydrological system is a low pass filter for all the HVs, and also a fractal filter for all the HVs but soil moisture. As a filter, the hydrological system has damping effect on the HVs:the longer the HVs progresses along the flow path, the smoother the time series, and the larger the temporal correlation. The sequence of their correlation strength is P< I< SF< ET< R< BF and SM< GL. The source of SF is complex and consists of precipitation, overland flow, interflow, and baseflow and thus its correlation strength are the combination of these processes. Moreover, the damping effect of the unsaturated zone is the strongest while that of the canopy and land surface is the weakest. In the Sagehen Creek watershed, the temporal variations of P and I are fractional Gaussion noise while ET, R, BF, and SF are fractional Brownian motion (fBm) and the temporal fluctuations of GL is second order fBm. The temporal correlation of GL decreases as the distance to the stream decreases due to the boundary effect of the stream. The hydrological responses, e.g., groundwater level and stream stages, have larger scaling exponents or stronger temporal correlation than the input signals, e.g., groundwater recharge and streamflow. |
PDF Full Text Request