Temporal Scale Analysis Of Catchment Hydrological Models Based On Stochastic Features

Catchment hydrological process takes on different patterns across temporal scales.To clarify the uncertainty revealed by observation and simulation during transition from the daily runoff generation evapotranspiration mechanism to annual water-heat correlation pattern is of fundamental importance in reaching a self-consistent observation simulation system.The stochastic soil moisture model provides reasonable solution to solve scale problems besides the classical paradigms of bottom-up and top-down approaches. Through incorporating the catchment storage capacity curve in the runoff generation portion of soil moisture stochastic equation, the thesis derived the function that describe basin-scale soil moisture dynamics. The derivation also suits for transforming other mechanism-focused conceptual hydrological models into probability form to clarify its dynamic properties.The ergodicity feature of stochastic soil moisture equation guarantees that the temporal average of soil moisture is the same as the ensemble average, thus, the long term water balance condition can be represented by the equilibrium solution of the stochastic function. The thesis simulated and analysed the scale and accuracy of applying its equilibrium solution in depicting the long range catchment hydrological pattern within the time and frequency domain. Results show that the soil moisture distribution can always reach its stable state given any meteorological and underlying surface conditions. Higher average soil moisture is correlated with more uniform underlying surface. The speed in reaching the equilibrium distribution is dominated by the relative magnitude of potential evapotranspiration. When focusing on small temporal scales, the soil moisture behaves as one step auto regress pattern in the time domain, red noise in the frequency domain.When focusing on large temporal scales, the soil moisture acts as stable process in the time domain, being white noise in the frequency domain, responsively. The water and energy supply determine the shape of the equilibrium distribution. The two factors determine the accuracy of applying the equilibrium distribution in depicting the long range catchment hydrological pattern through controlling the variance of the distribution. The variance is larger in catchments with higher water energy supplies.In order to quantify the uncertainty in observation and simulation across temporal scales, the thesis checks the significance of entropy and mutual information in a Bayesianview, quantized entropy of runoff observations can be used to represent the prior uncertainty in determining the catchment's hydrological patterns. Mutual information between runoff observation and the catchment's water energy provisions is employed to denote the uncertainty decrease given the existed observations. Mutual information between runoff observation and simulation is employed to denote the uncertainty decrease given the models. The differences of these items, as constrained by the functional transformation of the Bayes' theorem and data processing inequality, construct sound framework in evaluating the observation and simulation systems. An improved approach combining Knearest-neighbor method and support-vector-regression is employed to tackle with high dimensional information item estimation.We implement the information analysis with clustered daily hydrometeorological observations from MOPEX data set to analyse the uncertainty and its dominants across temporal scales. The estimations quantified the information contents and flows of hydrological items, the specific information contributions of former hydrological behaviours and new items. The estimations are closely related with the climate type of the catchments. It also shows that information distilled by the monthly and annual water balance models applied here does not correspond to that provided by observations around temporal scale from two months to half a year. This calls for a better understanding of seasonal hydrological mechanism.