Repository logo

Browsing by Author "Salas, Jose D., advisor"

Now showing 1 - 7 of 7
  • Thumbnail Image
    ItemOpen Access
    Disaggregation of precipitation records
    (Colorado State University. Libraries, 1991) Cadavid, Luis Guillermo, author; Salas, Jose D., advisor; Boes, Duane C., committee member; Yevjevich, Vujica M., 1913-, committee member; Fontane, Darrell G., committee member
    This investigation is related to temporal disaggregation of precipitation records. The objective is to formulate algorithms to disaggregate precipitation defined at a given time scale into precipitation of smaller time scales, assuming that a certain mechanism or stochastic process originates the precipitation process. The disaggregation algorithm should preserve the additivity property and the sample statistical properties at several aggregation levels. Disaggregation algorithms were developed for two models which belong to the class of continuous time point processes: Poisson White Noise (PWN) and Neyman-Scott White Noise (NSWN). Precipitation arrivals are controlled by a counting process and storm activity is represented by instantaneous amounts of precipitation (White Noise terms). Algorithms were tested using simulated samples and data collected at four precipitation stations in Colorado. The PWN model is the easiest and formulation of the disaggregation model was successful. The algorithm is based on the distribution of the number of arrivals (N) conditional on the total precipitation in the time interval (Y) , the distribution of the White Noise terms conditional on N and Y, and the distribution of the arrival times conditional on N. Its application to disaggregate precipitation is limited due to its lacl; of serial correlation. However, PWN disaggregation model performs well on PWN simulated samples. The NSWN is more complex. Required distributions are the same as for the PWN model. Formulation of a disaggregation algorithm was based on theoretical and empirical results. A procedure for model parameters estimation based on weighted least squares was implemented. This procedure reduces the number of estimation failures as compared to method of moments. NSWN disaggregation model performed well on simulated and recorded samples given that parameters used are similar to those controlling the process at the disaggregation scale. The main shortcoming is the incompatibility of parameter estimates at different aggregation levels. This renders the disaggregation model of limited application. Examination of variation of parameter estimates with the aggregation scale suggests the existence of a region where estimated values appear to be compatible. Finally, it is shown that the use of information at a nearby precipitation station with similar precipitation regime may improve parameter values to use in disaggregation.
  • Thumbnail Image
    ItemOpen Access
    Modeling the uncertainty of hydrologic processes exhibiting changes
    (Colorado State University. Libraries, 1998) Saada, Nidhal, author; Salas, Jose D., advisor; Yevjevich, Vujica M., 1913-, committee member; Boes, Duane C., committee member; Ramírez, Jorge A., committee member
    The Geometric-Normal-Normal (GNN) model was analyzed and tested for the purpose of simulating hydrologic processes that exhibit changes. The general moment equations of the GNN model were derived, particularly the lag-k autocorrelation function. They can be used to estimate the model parameters based on the method of moments. Other estimation methods were also suggested. They include regression analysis, fitting the autocorrelation function (ACF), using the range properties, and using the run properties. The performance of these methods was tested by using simulation experiments. The results showed that in terms of bias and mean square error the regression and range methods are better than the other methods for estimating the model parameters. The GNN model was applied to the White Nile River flows at Malakal and the annual net basin supply (NBS) data for Lake St. Clair of the Great Lakes system. Simulation experiments were conducted to test the ability of the GNN model to preserve a number of observed statistics such as the mean, standard deviation, skewness, rescaled range, Hurst coefficient, longest drought, maximum deficit, and surplus. Results show that the GNN model, in general, performs quite well in preserving these statistics. An extended version of the GNN model was also formulated and analyzed in this study. Different methods of estimation were suggested to estimate the model parameters. However, application of this model to Malakal flows and Lake St.Clair NBS data did not show any advantage over simpler GNN.
  • Thumbnail Image
    ItemOpen Access
    Stepwise nonparametric disaggregation for daily streamflow generation conditional on hydrologic and large-scale climatic signals
    (Colorado State University. Libraries, 2010) Molina Tabares, José Manuel, author; Ramírez, Jorge A., advisor; Salas, Jose D., advisor; Raff, David A., committee member; Kampf, Stephanie K., committee member
    A stepwise nonparametric stochastic disaggregation framework to produce synthetic scenarios of daily streamflow conditional on volumes of spring runoff and large-scale ocean-atmosphere oscillations is presented. This thesis examines statistical links (i.e., teleconnections) between decadal/interannual climatic variations in the Pacific Ocean and hydrologic variability in US northwest region, and includes a spectral analysis of climate signals to detect coherences of their behavior in the frequency domain. We explore the use of such teleconnections of selected signals (e.g., north Pacific gyre oscillation, southern oscillation, and Pacific decadal oscillation indices) in the proposed data-driven framework by means of a cross-validation-based combinatorial approach with the aim of simulating improved streamflow sequences when compared with disaggregated series generated from flows alone. A nearest neighbor time series bootstrapping approach is integrated with principal component analysis to resample from the empirical multivariate distribution. A volume-dependent scaling transformation is implemented to guarantee the summability condition. The downscaling process includes a two-level cascade scheme: seasonal-to-monthly disaggregation first followed by monthly-to-daily disaggregation. Although the stepwise procedure may lead to a lack of preservation of the historical correlation between flows of the last day of a month and flows of the first day of the following month, we present a new and simple algorithm, based on nonparametric resampling, that overcomes this limitation. The downscaling framework presented here is parsimonious in parameters and model assumptions, does not generate negative values, and preserves very well the statistical characteristics, temporal dependences, and distributional properties of historical flows. We also show that both including conditional information of climatic teleconnection signals and developing the downscaling in cascades decrease significantly the mean error between synthetic and observed flow traces. The downscaling framework is tested with data from the Payette River Basin in Idaho.
  • Thumbnail Image
    ItemOpen Access
    Stochastic characterization of droughts in stationary and periodic series
    (Colorado State University. Libraries, 2008) Cancelliere, Antonino, author; Salas, Jose D., advisor; Boes, Duane C., advisor
    Stochastic modelling of droughts is a topic of great interest in water resources management. For instance, estimating drought probabilities and return periods helps in implementing risk based management decisions of water supply systems and provide useful information for drafting drought management plans. Due to the limited number of droughts that can be generally observed in historical series, the inferential approach, e.g. fitting a probability distribution to drought characteristics from an observed hydrological sample, leads to unreliable results. Furthermore, the multiyear spanning of droughts, as well as their multivariate framework requires the development of concepts and tools that differ significantly from those generally adopted to analyze other hydrological extremes, such as floods.
  • Thumbnail Image
    ItemOpen Access
    Stochastic modeling of seasonal streamflow
    (Colorado State University. Libraries, 1987) Mendonça, Antonio Sergio Ferreira, author; Salas, Jose D., advisor; Fontane, Darrell G., committee member; Loftis, Jim C., committee member; Gessler, Johannes, committee member
    This research examines topics on seasonal (monthly, bimonthly, etc.) hydrologic time-series modeling. A family of periodic models was derived by allowing parameters for a particular Multiplicative Autoregressive Integrated Moving Average model (Multiplicative ARIMA) to vary from season to season. The derived model presents parameters relating data for seasons in the same year and parameters relating data for the same season for consecutive years. PARMA models are particular cases of the proposed model, here called Multiplicative Periodic Autoregressive Moving Average (Multiplicative PARMA). Least-squares estimation based on the Powell algorithm for nonlinear optimization was developed for determining the model parameters. Properties such as seasonal variances and autocorrelations were derived analytically for particular cases of the general model. Analysis of sensitivity of the annual autocorrelograms to the parameters of the model showed that the yearly autoregressive parameters are the most important for the reproduction of high annual autocorrelations. Tests of model were made through data generation. The model was applied to four-and six-season series for river discharge presenting distinct characteristics of variabilty and dependence. Tests for goodness-of-fit and selection criteria of models for seasonal series were also discussed. Results from data generation indicate that the estimation procedure is able to estimate parameters for the Multiplicative PARMA models and can also be used for refinement of estimations made by method-of-moments for other models. Application to discharge data from St. Lawrence, Niger, Elkhorn and Yellowstone rivers showed that the proposed modeling technique is able to preserve long term dependence better than models currently used in practical hydrology. Direct consequence of this improvement is better reproduction of floods and droughts and more accuracy in the design and operation of water resource structures.
  • Thumbnail Image
    ItemOpen Access
    Stochastic simulation of hydrologic data based on nonparametric approaches
    (Colorado State University. Libraries, 2008) Lee, Taesam, author; Salas, Jose D., advisor
    Stochastic simulation of hydrologic data has been widely developed for several decades. However, despite the several advances made in literature still a number of limitations and problems remain. The major research topic in this dissertation is to develop stochastic simulation approaches to tackle some of the existing problems such as the preservation of the long-term variability and the joint modeling of intermittent and non-intermittent stations. For this purpose, nonparametric techniques have been applied. For simulating univariate seasonal streamflows, a model is suggested based on k-nearest neighbors resampling (KNNR). Gamma kernel density estimate (KDE) perturbation is employed to generate realistic values of streamflow that are not part of the historical data. Further, aggregate and pilot variables are included in KNNR so as to reproduce the long-term variability. For multivariate streamflows, the moving block bootstrapping procedure is employed considering a random block length, KNNR block selection to avoid the discontinuity between blocks, a Genetic Algorithm mixture, and Gamma KDE perturbation. In addition, the drawbacks of an existing nonparametric disaggregation scheme have been examined and appropriate modifications developed that include accurate adjusting for the disaggregate variable, KNNR, and Genetic Algorithm mixture. The suggested univariate, multivariate, and disaggregation models have been compared with existing nonparametric models using several cases of streamflow data of the Colorado River System. In all cases, the results showed major improvements. Furthermore, disaggregation from daily to hourly rainfall for a single site has been studied based on three disaggregation models so as to account for the diurnal cycle in hourly data. Those models are (1) Conditional Markov Chain and Simulated Annealing (CMSA), (2) Product Model (GAR(1)-PDAR(1)) with Accurate Adjusting (PGAA), and (3) Stochastic Selection Method with Weighted Storm Distribution (SSMW). Various tests and comparisons have been performed to validate the models and it revealed that PGAA is superior to the others for preserving the diurnal cycle and the key statistics of hourly rainfall.
  • Thumbnail Image
    ItemOpen Access
    The effect of parameter uncertainty in stochastic streamflow simulation
    (Colorado State University. Libraries, 2009) Lee, Dong-Jin, author; Salas, Jose D., advisor
    Hydrologic time series simulation based on a stochastic model is intended to obtain a set of equally likely hydrologic sequences that could possibly occur in the future and might be useful for determining the uncertainty of decision variables such as the storage capacity of a reservoir. Since stochastic models generally hinge on parameters that are estimated based on a limited historical sample, the model parameters become uncertain and so are any decision variables that are derived from the generated samples. The main objective of this study is to propose and analyze methods for quantifying the effect of parameter uncertainty of the models that are used in the generation of synthetic streamflow series. As a way of quantifying parameter uncertainty of a stochastic model, asymptotic and Bayesian approaches have been implemented and their performances compared through extensive simulation experiments. Alternative streamflow simulation techniques have been utilized with parameter uncertainty incorporated such as stochastic models of annual streamflows at single and multiple sites as well as temporal and spatial disaggregation models. The impact of parameter uncertainty is shown to increase the variability of generated flow statistics and resultant design related variables, which is visible even with a relatively large sample size, e.g. sample size of 200. The Bayesian approach produces larger variability of generated statistics for small sample sizes than the asymptotic approach, and the difference between the two approaches is more evident for the case of generation of streamflows with high serial correlations. The effect of parameter uncertainty within disaggregation models is not as significant on the first and second moments of disaggregated flows as the effect of parameter uncertainty of the models that generate the input variables; whereas the effect of parameter uncertainty of disaggregation models results in more variability of month-to-month, month-to-annual, and cross correlations than those induced by the uncertainty of the model parameters of input variables.