A major area targeted for hydrometeorological forecast service improvements is in flash flood forecasting. Verification data show that general public service products of flash flood forecasts do not provide enough lead time in order for the public to make effective response. Sophisticated users of flash flood forecasts could use forecast probabilities of flash flooding in order to make decisions in preparation for the predicted event. To this end, a systematic probabilistic approach to flash flood forecasting is presented. The work first describes a deterministic system which serves as a conceptual basis for the probability system. The approach uses accumulated rainfall plus potential rainfall over a specified area and time period, and assesses this amount against the water holding capacity of the affected basin. These parameters are modeled as random variables in the probabilistic approach. The effects of uncertain measurements of rainfall and forecasts of precipitation from multiple information sources within a time period and moving forward in time are resolved through the use of Bayes' Theorem. The effect of uncertain inflows and outflows of atmospheric moisture on the states of the system, the transformation of variables, is resolved by use of convolution. Requirements for probability distributions to satisfy Bayes' Theorem are discussed in terms of the types and physical basis of meteorological data needed. The feasibility of obtaining the data is evaluated. Two alternatives for calculating the soil moisture deficit are presented--one, an online automatic rainfall/runoff model, the other an approximation. Using the soil moisture approximation, a software program was developed to test the probabilistic approach. A storm event was simulated and compared against an actual flash flood event. Results of the simulation improved forecast lead time by 3-5 hours over the actual forecasts issued at the time of the event.

Models used as part of quantitative studies of vadose zone processes are becoming increasingly complex. However, even the most elaborate models can not capture the complex interactions between spatially distributed water, plant, and atmospheric components of the unsaturated flow system. These processes will always need to be approximated by relatively simple mathematical expressions with limited parameterization. Because of this, there is an ever increasing awareness among hydrologists of the need to describe and quantify these uncertainties to better understand the utility of model predictions and inform decisions concerning model development and data collection. Significant developments in the most recent generation of parameter estimation codes have facilitated the estimation of parameters and quantification of the associated uncertainty in the parameter estimates and model structure; however, these codes are computationally expensive. To facilitate the proposed analysis of more computationally efficient models are required.Computationally efficient models do not necessarily imply over simplified models In the appropriate context, simplifications are possible that reduce the complexity of the model but do not reduce the complexity of the system being represented by the model. I investigate a series of approaches to reduce the computational load of models, facilitating inverse analysis with readily available computing facilities.In light of the improvements to the methodology of parameter estimation, the success of the analysis still depends on the observed response to which the model is compared; the data and the information contained in the data. Given limited resources (both cost and technology) it is important to identify those data that will provide the greatest information about a system. To this end, the investigations presented here also investigate methods to identify informative data and to extract information from data effectively.

The weight ratio of naturally occurring chloride and bromide (Cl/Br) can be used to identify water origin, and as a tracer. In the Tucson basin, where geologic inputs of these constituents are limited, the mean Cl/Br ratio of uncontaminated ground water is 130. This ratio may be altered by artificial recharge, sewage effluent disposal, and agricultural practices. In this study, Cl/Br ratios of 30 Tucson urban runoff samples averaged 34, one quarter that of the ground water. Cl/Br ratios may be useful in studying the fate of urban runoff recharged via dry wells. Urban runoff contains bromide from ethylene dibromide in leaded gasoline. Both species behaved conservatively in sorption and volatilization experiments; thus, these processes should not significantly alter Cl/Br ratios in surface or ground water. Bromide analytical methods were compared and methods best suited for analysis of dilute natural waters are discussed.

A two-phase study of the Upper Santa Cruz River Basin consisting of a hydrologic assessment and the application of a computer model to the area is presented. Groundwater occurs in Older and Younger Alluvium units, is held under unconfined conditions, and is of good quality. The Finite-Difference Three-Dimensional Groundwater Flow Model (MODFLOW) is applied to the study area in order to simulate the hydrologic system. A water budget and conceptual model is developed to aid in model input and calibration. A steady-state analysis for the system is performed for 1965 in order to calibrate model parameters and produce initial conditions for the transient analysis. A transient analysis is performed to verify model response over the period 1965-1980 in which stresses to the system are changing. Model-produced trends and water levels are consistent with observed data, indicating the model's utility for predictive analysis of the system.

Three hydrologic simulation models of different resolutions were evaluated to determine model response to predicting runoff under changing vegetation cover. Two empirically-based regression models (Baker-Kovner Streamflow Regression Model and ECOSIM) and one multiple component water balance model (Yield) were modified, using FORTRAN 77 and calibrated on a southwestern ponderosa pine ecosystem. Statistical analysis indicate no significant difference between the Baker-Kovner and Yield models, while ECOSIM consistently under predicts by as much as 50 percent from the observed runoff. This is mainly attributed to a sensitivity to the insolation factor. Yield is the best predictor for moderate and high flows, to within 10 and 20 percent respectively. Of the four watershed treatments, the light overstory thinning on Watershed 8 yielded the best response for all three models. This is in contrast to the strip-cut treatment on Watershed 14 which consistently over-predicted, in large part due an inaccurate estimation of snowpack evaporation on the exposed, south-facing strip-cuts. Runoff responses are highly influenced by the precipitation regime, soil and topographic characteristics of a watershed as well as by a reduction in evapotranspiration losses from changes in vegetation cover.

Measuring the difference between the measured hydrocarbon thickness in an observation well and the actual hydrocarbon thickness in a formation has been a major hinderance in estimating the volume of free product in aquifers. The thickness of hydrocarbons in some observation wells is as much as four times greater than in the formation. A three-dimensional sandbox model was used to determine linear relationships between the height of hydrocarbon in an observation well to the thickness of hydrocarbon above the water table. A fine-grained sand was used to calibrate the system, then two soils were characterized and modeled. The results of the sand and natural soil runs were compared with three models found in the literature. The linear relationships developed from the sandbox experiments compared favorably with the models in the literature. Further investigation with a more accurate petroleum thickness detection method will result in site-specific linear relationships.

Two regional aquifers are analyzed based on their stable isotope and hydrochemical data. The Tucson basin (Arizona) and Madrid basin (Spain), present similarities in their geologic patterns. The ground-water flow regimens, however, are different. Madrid basin presents local, intermediate and regional flow. Rivers in the Madrid basin are perennial, while Tucson washes only drain surface run-off. The winter precipitation is the main source for recharge in both basins. This was deduced by the difference in oxygen-18 content between winter and summer precipitation over both basins. Madrid basin presents a continental effect occurring from west to east and an altitude effect from northeast to southwest. Tucson basin presents an altitude effect. The oxygen-18 content decreases toward the higher elevations. Mixing of waters in discharge areas in both basins due to anthropogenic activity masked relationships, if any, among the stable isotopes and hydrochemical species.

The chemistry of ground water from the Lower San Pedro River basin between Reddington and Mammoth, Arizona, was studied. Objectives were to determine (1) recharge and discharge relationships among aquifers through a comparison of their water chemistries and (2) whether copper mill tailings at San Manuel have affected ground-water quality. The study area has four aquifers: (1) the Catalina aquifer along the basin's western flank, (2) the Galiuro aquifer along the eastern flank, (3) the artesian aquifer, and (4) the channel aquifer. The latter two are in the center of the basin. This study produced the following conclusions: Ground water from the Catalina and Galiuro aquifers recharge the artesian aquifer. The age of ground water in the artesian aquifer is between 7,700 and 13,800 years. The channel aquifer is recharged by the artesian aquifer and intermittent surface flows. The chemistry of water in the channel aquifer is largely controlled by the dissolution of evaporites. There is no evidence that the mill tailings at San Manuel have affected ground-water quality.

The effectiveness of soil aquifer treatment (SAT) to renovate both tertiary (Site 1) and secondary (Site 2) wastewater was investigated using two small recharge basins. Suction lysimeters collected soil water samples at various depths within the vadose zone from 0.5 to 20.0 feet. Total organic halide (TOX) and dissolved organic carbon (DOC) were evaluated to determine SAT efficiency in removing TOX and DOC found in the source waters. Approximately 77% of the TOX was removed from the tertiary wastewater during percolation through the vadose zone at Site 1, while the secondary source water underwent a maximum reduction of approximately 55% at Site 2. At Site 1 there was mobilization of in-situ humic materials causing increases in DOC concentrations at depth. These decreased with time and after 83 hours DOC concentrations at 20 feet were 51% of inflow concentration. At Site 2 DOC removal ranged from 44 to 73% at a depth of 15 feet for three flooding events. Discrepancies between the two locations are thought to reflect lithologic differences and dissimilarities in hydraulic loading rates.

The Rancho California Water District has an opportunity to conjunctively manage available ground, surface, imported, and reclaimed water resources through a program of artificial recharge and recovery. The site of proposed operations is eastern Pauba Valley, Riverside County, California. Hydrogeological analyses were conducted to identify valley aquifers, evaluate hydraulic properties and quantify ground water storage, movement and yield. Field studies included nine aquifer tests, a survey of well locations and construction of five monitoring wells. All data collected were entered into a computerized Water Resources Information Management System (WRIMS) custom configured for the project. A finite-difference computer simulation model was constructed to evaluate the feasibility and facilities requirements of two alternative recharge and recovery programs. Computer simulation results confirm the feasibility of 9,000 and 18,000 acre feet per year recharge and recovery programs.