Les systèmes de prévision des crues décrivent les transformations pluie-débit en se basant sur des représentations simplifiées. Ces représentations modélisent les processus physiques impliqués avec des descriptions empiriques, ou basées sur des équations de la mécanique classique. Les performances des modèles actuels de prévision des crues sont affectées par différentes incertitudes liées aux approximations et aux paramètres du modèle, aux données d’entrée et aux conditions initiales du bassin versant. La connaissance de ces incertitudes permet aux décideurs de mieux interpréter les prévisions et constitue une aide à la décision lors de la gestion de crue. L’analyse d’incertitudes dans les modèles hydrologiques existants repose le plus souvent sur des simulations de Monte-Carlo (MC). La mise en œuvre de ce type de techniques requiert un grand nombre de simulations et donc un temps de calcul potentiellement important. L'estimation des incertitudes liées à la modélisation hydrologique en temps réel reste donc une gageure. Dans ce projet de thèse, nous développons une méthodologie de prévision des crues basée sur les réseaux Bayésiens (RB). Les RBs sont des graphes acycliques dans lesquels les nœuds correspondent aux variables caractéristiques du système modélisé et les arcs représentent les dépendances probabilistes entre ces variables. La méthodologie présentée propose de construire les RBs à partir des principaux facteurs hydrologiques contrôlant la génération des crues, en utilisant à la fois les observations disponibles de la réponse du système et les équations déterministes décrivant les processus concernés. Elle est conçue pour prendre en compte la variabilité temporelle des différentes variables impliquées. Les dépendances probabilistes entre les variables (paramètres) peuvent être spécifiées en utilisant des données observées, des modèles déterministes existants ou des avis d’experts. Grâce à leurs algorithmes d’inférence, les RBs sont capables de propager rapidement, à travers le graphe, différentes sources d'incertitudes pour estimer leurs effets sur la sortie du modèle (ex. débit d'une rivière). Plusieurs cas d’études sont testés. Le premier cas d’étude concerne le bassin versant du Salat au sud-ouest de la France : un RB est utilisé pour simuler le débit de la rivière à une station donnée à partir des observations de 3 stations hydrométriques localisées en amont. Le modèle présente de bonnes performances pour l'estimation du débit à l’exutoire. Utilisé comme méthode inverse, le modèle affiche également de bons résultats quant à la caractérisation de débits d’une station en amont par propagation d’observations de débit sur des stations en aval. Le deuxième cas d’étude concerne le bassin versant de la Sagelva situé en Norvège, pour lequel un RB est utilisé afin de modéliser l'évolution du contenu en eau de la neige en fonction des données météorologiques disponibles. Les performances du modèle sont conditionnées par les données d’apprentissage utilisées pour spécifier les paramètres du modèle. En l'absence de données d'observation pertinentes pour l’apprentissage, une méthodologie est proposée et testée pour estimer les paramètres du RB à partir d’un modèle déterministe. Le RB résultant peut être utilisé pour effectuer des analyses d’incertitudes sans recours aux simulations de Monte-Carlo. Au regard des résultats enregistrés sur les différents cas d’études, les RBs se révèlent utiles et performants pour une utilisation en support d’un processus d'aide à la décision dans le cadre de la gestion du risque de crue. / Flood forecasting describes the rainfall-runoff transformation using simplified representations. These representations are based on either empirical descriptions, or on equations of classical mechanics of the involved physical processes. The performances of the existing flood predictions are affected by several sources of uncertainties coming not only from the approximations involved but also from imperfect knowledge of input data, initial conditions of the river basin, and model parameters. Quantifying these uncertainties enables the decision maker to better interpret the predictions and constitute a valuable decision-making tool for flood risk management. Uncertainty analysis on existing rainfall-runoff models are often performed using Monte Carlo (MC)- simulations. The implementation of this type of techniques requires a large number of simulations and consequently a potentially important calculation time. Therefore, quantifying uncertainties of real-time hydrological models is challenging. In this project, we develop a methodology for flood prediction based on Bayesian networks (BNs). BNs are directed acyclic graphs where the nodes correspond to the variables characterizing the modelled system and the arcs represent the probabilistic dependencies between these variables. The presented methodology suggests to build the RBs from the main hydrological factors controlling the flood generation, using both the available observations of the system response and the deterministic equations describing the processes involved. It is, thus, designed to take into account the time variability of different involved variables. The conditional probability tables (parameters), can be specified using observed data, existing hydrological models or expert opinion. Thanks to their inference algorithms, BN are able to rapidly propagate, through the graph, different sources of uncertainty in order to estimate their effect on the model output (e.g. riverflow). Several case studies are tested. The first case study is the Salat river basin, located in the south-west of France, where a BN is used to simulate the discharge at a given station from the streamflow observations at 3 hydrometric stations located upstream. The model showed good performances estimating the discharge at the outlet. Used in a reverse way, the model showed also satisfactory results when characterising the discharges at an upstream station by propagating back discharge observations of some downstream stations. The second case study is the Sagelva basin, located in Norway, where a BN is used to simulate the accumulation of snow water equivalent (SWE) given available weather data observations. The performances of the model are affected by the learning dataset used to train the BN parameters. In the absence of relevant observation data for learning, a methodology for learning the BN-parameters from deterministic models is proposed and tested. The resulted BN can be used to perform uncertainty analysis without any MC-simulations to be performed in real-time. From these case studies, it appears that BNs are a relevant decisionsupport tool for flood risk management.

Bibliography: leaves [175-181]

Les crues « éclair » parfois dévastatrices qui touchent les bassins versants méditerranéens du Sud de la France sont difficiles à anticiper. Leur prévision passe par l'utilisation de modèles pluie-débit, dont l'efficacité est encore limitée par les incertitudes liées notamment à la variabilité spatiale des pluies méditerranéennes et à la caractérisation de l'état hydrique initial des hydrosystèmes. Dans le cas de bassins karstiques, à ces incertitudes s'ajoutent celles liées à la dynamique des aquifères et à leur rôle sur la formation des crues. La première partie de ce travail de thèse propose un modèle pluie-débit horaire, distribué, événementiel et parcimonieux pour reproduire les crues « éclair » à l'exutoire du bassin karstique du Lez (Montpellier) de 114 km2. Le modèle est évalué non seulement sur la qualité des simulations de débits mais aussi sur la qualité de son initialisation obtenu grâce à une relation entre sa condition initiale et divers indicateurs de l'état hydrique de l'hydrosystème. Calibré sur 21 épisodes de crues, le modèle fournit des simulations satisfaisantes, et sa condition initiale est significativement corrélée à l'indice d'humidité Hu2 du modèle SIM de Météo-France ou à la piézométrie dans l'aquifère du Lez. Les pluies mesurées par radar en début d'automne sont de bonne qualité et conduisent à une amélioration des simulations de débit et de l'estimation de la condition initiale du modèle. En revanche, les pluies mesurées par radar en fin d'automne sont de moindre qualité et n'améliorent pas les simulations. Face aux incertitudes liées à la paramétrisation du modèle ou à l'estimation des pluies radar, la deuxième partie du travail de thèse analyse l'apport de l'assimilation des débits observés pour corriger en temps réel les paramètres les plus sensibles du modèle et notamment sa condition initiale ou les pluies radar en entrée du modèle. La procédure d'assimilation de données a été mise en place à l'aide du coupleur PALM, qui permet de relier modèle hydrologique à l'algorithme d'assimilation. La correction de la condition initiale du modèle permet généralement d'améliorer les prévisions (sous hypothèse de pluie future connue); la correction de la pluie a des effets similaires. Néanmoins les limites de cette correction sont atteintes dans le cas où le modèle ne reproduit pas de façon satisfaisante la partie initiale de montée des eaux, ce qui pourra être amélioré par la suite. Finalement, ce travail de thèse montre que la complexité d'un bassin karstique peut être représentée efficacement à l'aide d'un nombre réduit de paramètres, pour simuler les débits, et contribue à l'amélioration des outils opérationnels pour la prévision des crues. / The sometimes devastating flash floods which affect the Mediterranean watersheds of the South of France are difficult to anticipate. Flood forecasting requires the use of rainfall-runoff models which are limited in their efficiency by uncertainty related to the spatial variability of Mediterranean rainfall and the characterization of the initial hydric state of the system. In karstic catchments, these uncertainties are added to those due to aquifer dynamics and their role in flood genesis. The first part of this work will present a distributed event-based parsimonious hourly rainfall-runoff model in order to reconstruct flash flood events at the outlet of the 114 km2 Lez Catchment (Montpellier). The model is evaluated not only for the quality of the simulations produced, but for the quality of its parameter initialization obtained using a relationship between the initial condition and various hydric state indicators of the system. Calibrated using 21 flood episodes, the model produces satisfactory simulations and its initial condition is significantly correlated with the Hu2 soil humidity index of the Météo-France model or piezometers measuring the Lez aquifer. Radar rainfall data measured in early fall are of good quality and lead to improved discharge simulations and an improved estimation of the model initial condition. However, rainfall measured by radar in late fall are of poor quality and do not improve the simulations. Confronted with the uncertainty related to model parametrization or the estimation of radar rainfall, the second part of this dissertation analyzes improvements achieved by assimilating observed discharge measurements in order to perform real-time corrections to the most sensitive model parameters and notably the initial condition and the radar rainfall input to the model. The data assimilation procedure was implemented with the help of the PALM coupling software which allows for the linking of the hydrological model with the assimilation algorithm. Correcting the initial condition allowed for, on average, the improvement of forecasting (under a known future rainfall hypothesis); correcting the rainfall had similar effects. Nevertheless, the limits of this approach are reached when the model is unable to satisfactorily reproduce the rising limb of the hydrograph, a problem which may be addressed by future research. Finally, this body of work demonstrates that the complexity of a karstic catchment can be efficiently represented with a reduced number of parameters in order to simulate discharges and contribute to the improvement of operational tools for flood forecasting.

A detailed and comprehensive description of the state of the art in the field of flood forecasting opens this work. Advantages and shortcomings of currently available methods are identified and discussed. Amongst others, one important aspect considers the most exigent weak point of today’s forecasting systems: The representation of all the fundamentally different event specific patterns of flood formation with one single set of model parameters. The study exemplarily proposes an alternative for overcoming this restriction by taking into account the different process characteristics of flood events via a dynamic parameterisation strategy. Other fundamental shortcomings in current approaches especially restrict the potential for real time flash flood forecasting, namely the considerable computational requirements together with the rather cumbersome operation of reliable physically based hydrologic models. The new PAI-OFF methodology (Process Modelling and Artificial Intelligence for Online Flood Forecasting) considers these problems and offers a way out of the general dilemma. It combines the reliability and predictive power of physically based, hydrologic models with the operational advantages of artificial intelligence. These operational advantages feature extremely low computation times, absolute robustness and straightforward operation. Such qualities easily allow for predicting flash floods in small catchments taking into account precipitation forecasts, whilst extremely basic computational requirements open the way for online Monte Carlo analysis of the forecast uncertainty. The study encompasses a detailed analysis of hydrological modeling and a problem specific artificial intelligence approach in the form of artificial neural networks, which build the PAI-OFF methodology. Herein, the synthesis of process modelling and artificial neural networks is achieved by a special training procedure. It optimizes the network according to the patterns of possible catchment reaction to rainstorms. This information is provided by means of a physically based catchment model, thus freeing the artificial neural network from its constriction to the range of observed data – the classical reason for unsatisfactory predictive power of netbased approaches. Instead, the PAI-OFF-net learns to portray the dominant process controls of flood formation in the considered catchment, allowing for a reliable predictive performance. The work ends with an exemplary forecasting of the 2002 flood in a 1700 km² East German watershed.

Aus der Einleitung: The calculation and prediction of river flow is a very old problem. Especially extremely high values of the runoff can cause enormous economic damage. A system which precisely predicts the runoff and warns in case of a flood event can prevent a high amount of the damages. On the basis of a good flood forecast, one can take action by preventive methods and warnings. An efficient constructional flood retention can reduce the effects of a flood event enormously.With a precise runoff prediction with longer lead times (>48h), the dam administration is enabled to give order to their gatekeepers to empty dams and reservoirs very fast, following a smart strategy. With a good timing, that enables the dams later to store and retain the peak of the flood and to reduce all effects of damage in the downstream. A warning of people in possible flooded areas with greater lead time, enables them to evacuate not fixed things like cars, computers, important documents and so on. Additionally it is possible to use the underlying rainfall-runoff model to perform runoff simulations to find out which areas are threatened at which precipitation events and associated runoff in the river. Altogether these methods can avoid a huge amount of economic damage.:List of Symbols and Abbreviations S. III 1 Introduction S. 1 2 Process based Rainfall-Runoff Modelling S. 5 2.1 Basics of runoff processes S. 5 2.2 Physically based rainfall-runoff and hydrodynamic river models S. 15 3 Portraying Rainfall-Runoff Processes with Neural Networks S. 21 3.1 The Challenge in General S. 22 3.2 State-of-the-art Approaches S. 24 3.3 Architectures of neural networks for time series prediction S. 26 4 Requirements specification S. 33 5 The PAI-OFF approach as the base of the system S. 35 5.1 Pre-Processing of the Input Data S. 37 5.2 Operating and training the PoNN S. 47 5.3 The PAI-OFF approach - an Intelligent System S. 52 6 Design and Implementation S. 55 6.1 Design S. 55 6.2 Implementation S. 58 6.3 Exported interface definition S. 62 6.4 Displaying output data with involvement of uncertainty S. 64 7 Results and Discussion S. 69 7.1 Evaluation of the Results S. 69 7.2 Discussion of the achieved state S. 75 8 Conclusion and FutureWork S. 77 8.1 Access to real-time meteorological input data S. 77 8.2 Using further developed prediction methods S. 79 8.3 Development of a graphical user interface S. 80 Bibliography S. 83

Thesis (M.Sc. (Geography)) -- University of Limpopo, 2019 / Floods are one of the major natural hazards that occur with devastating effects globally. South Africa is one of the countries that is affected as flooding frequently occurs at different sub-national scales and with devastating impacts on human settlements. The variability of the nature, impact and frequency of flood occurrence in the country has heightened interest in the assessment and determination of flooding vulnerability, particularly in areas that have been affected or are likely to be affected in the near future. Given the uncertainties surrounding flood occurrence and the enormous damages resulting from the events, this study sought to assess the vulnerability of settlements to floods in Ward 7 and 9 of Lephalale Local Municipality. To accomplish this, both primary and secondary sources of data were used in this study. A mixture of closed-ended and open-ended household questionnaire, which was administered to 133 and 227 randomly selected households in Ward 7 and 9 respectively was used. In addition, a vulnerability index was developed using an indicator approach in order to determine levels of flood vulnerability in the study areas. Indicators were identified, grouped and normalised using the standardization method, then weighted using pairwise comparison method. The various indicators were then aggregated through a linear summation method into a vulnerability index. This index was subsequently used to produce a vulnerability map showing the spatial pattern of the different flood vulnerability levels in the studied areas. The results reveal that socioeconomic as well as physical factors influence settlements’ vulnerability to flooding disasters. Furthermore, the vulnerability index map showed that Ward 7 was more vulnerable to flooding, with an average index of about 0.16 while Ward 9 was less vulnerable, with an average flood vulnerability index of 0,07. The vulnerability map also indicated that out of the total land area of 13.54km2 occupied by settlements in Ward 7, 9.38 km2 was very vulnerable, 2.27km2 highly vulnerable and 1.89km2 had low vulnerability. In Ward 9, about 4.44km2 of settlements land was experiencing low vulnerability while 29.96km2 experienced very low levels of vulnerability. The study concludes that the high vulnerability of Ward 7 was a result of an interplay of factors that include its nearness to the stream, a high proportion of low-lying land, land use type and high population densities. The results of this study can serve as a basis for targeting prioritization efforts, emergency response measures, and policy interventions at the ward level for minimizing flood disaster vulnerability in municipal areas. The study recommends that flood vulnerability assessments should integrate socio-economic characteristics with physical factors in order to adequately assess vulnerability and therefore enable municipalities to anticipate floods and plan for them. / University of Limpopo Staff Financial Assistance, Risk and Vulnerability Science Centre and VLIR

Flooding is one of the deadliest natural hazards in the world. Forecasting floods in advance can significantly reduce the socio-economic impacts. An accurate and reliable flood forecasting system is heavily dependent on the input precipitation data. Real-time, spatially, and temporally continuous Radar Quantitative Precipitation Estimates (QPEs) is useful precipitation information source. This research aims to investigate the efficacy of American and Canadian weather radar QPEs on hydrological model calibration and validation for flood forecasting in urban and semi-urban watersheds in Canada. A comprehensive review was conducted on the weather Radar network and its’ hydrological applications, challenges, and potential future research in Canada. First, radar QPEs were evaluated to verify the reliability and accuracy as precipitation input for hydrometeorological models. Then, the radar-gauge merging techniques were assessed to select the best method for urban flood forecasting applications. After that, merged Radar QPEs were used as precipitation input for the hydrological models to assess the impact of radar QPEs on hydrological model calibration and validation. Finally, a framework was developed by integrating hydrological and hydraulic models to produce flood forecasts and inundation maps in urbanized watersheds. Results indicated that dual-polarized radar QPEs could be effectively used as a source of precipitation input to hydrological models. The radar-gauge merging enhances both the accuracy and reliability of Radar QPEs, and therefore, the accuracy of streamflow simulation is also improved. Since flood forecasting agencies usually use hydrological models calibrated and validated using gauge data, it is recommended to use bias-corrected Radar QPEs to run existing hydrological models to simulate streamflow to produce flood extent maps. The hydrological and hydraulic models could be integrated into one framework using bias-corrected Radar QPEs to develop a successful flood forecasting system. / Thesis / Doctor of Science (PhD) / Floods are common and increasing deadly natural hazards in the world. Predicting floods in advance using Flood Early Warning System (FEWS) can facilitate flood mitigation. Radar Quantitative Precipitation Estimates (QPEs) can provide real-time, spatially, and temporally continuous precipitation data. This research focuses on bias-correcting and evaluating radar QPEs for hydrologic forecasting. The corrected QPE are applied into a framework connecting hydrological and hydraulic models for operational flood forecasting in urban watersheds in Canada. The key contributions include: (1) Dual-polarized radar QPEs is a useful precipitation input to calibrate, validate and run hydrological models; (2) Radar-gauge merging enhance accuracy and reliability of radar QPEs; (3) Floods could be more accurately predicted by integrating hydrological and hydraulic models in one framework using bias-corrected Radar QPEs; and (4) Gauge-calibrated hydrological models can be run effectively using the bias-corrected radar QPEs. This research will benefit future applications of real-time radar QPEs in operational FEWS.

Many techniques have been developed over the years in first world countries for the estimation of flood hydrographs from small catchments for application in design, management and operations of water related issues. However, relatively little attention has been directed towards the transfer and adaptation of such techniques to developing countries in which major hydrological decisions are crucially needed, but in which a scarcity of quality hydrological data often occurs. As a result, hydrologists and engineers in developing countries are frequently unable to alleviate the problems that extreme rainfall events can create through destructive flood flows or, alternatively, they do not possess the appropriate tools with which to design economically viable hydraulic structures. Eritrea is a typical example of a developing country which faces difficulties in regard to the adaptation of an appropriate design flood estimation technique for application on small catchments. As a result, the need has arisen to adapt a relatively simple and robust design flood model that can aid hydrologists and engineers in making economic and safe designs of hydraulic structures in small catchments. One objective of this study was, therefore, to review approaches to hydrological modelling and design flood estimation techniques on small catchments, in order to identify the barriers regarding their adaptation, as well as to assist in the selection of an appropriate technique for application, in Eritrea. The southern African adaptation of the SCS (i.e. Soil Conservation Service) design hydrograph technique, which has become a standard method for design flood estimation from small catchments in that region, was selected for application on small catchments in Eritrea for several reasons. It relies on the determination of a simple catchment response index in the form of an initial Curve Number (CN), which reflects both the abstraction characteristics and the non-linear stormflow responses of the catchment from a discrete rainfall event. Many studies on the use of SCS-based hydrological models have identified that adjustment of the initial CN to a catchment's antecedent soil moisture (ASM) to be crucial, as the ASM has been found to be one of the most sensitive parameters for accurate estimates of design flood volumes and peak discharges. In hydrologically heterogeneous regions like Eritrea, the hypothesis was postulated that simulations using a suitable soil water budgeting procedure for CN adjustment would lead to improved estimates of design flood volumes and peak discharges when compared with adjustments using the conventional SCS antecedent moisture conditions (SCS-AMC) method. The primary objective of this dissertation was to develop a surrogate methodology for the soil water budgeting procedure of CN adjustment, because any direct applications of soil water budgeting techniques are impractical in most parts of Eritrea owing to a scarcity of adequate and quality controlled hydrological information. It was furthermore hypothesised that within reasonably similar climatic regions, median changes in soil moisture storage from the socalled "initial" catchment soil moisture conditions, i.e. LIS, were likely to be similar, while between different climatic regions median LISs were likely to be different. Additionally, it was postulated that climatic regions may be represented by a standard climate classification system. Based on the above hypotheses, the Koppen climate classification, which can be derived from mean monthly rainfall and temperature information, was first applied to the 712 relatively homogeneous hydrological response zones which had previously been identified in southern Africa. A high degree of homogeneity of median values of LIS, derived by the daily time step ACRU soil moisture budgeting model, was observed for zones occurring within each individual Koppen climate class (KCC) - this after a homogeneity test had been performed to check if zones falling in a specific KCC had similar values of median LIS. Further assessment within each KCC found in southern Africa then showed that a strong relationship existed between LIS and Mean Annual Precipitation (MAP). This relationship was, however, different between KCCs. By developing regression equations, good simulations of median LIS from MAP were observed in each KCC, illustrating the potential application of the Koppen climate classification system as an indicator of regional median LIS, when only very basic monthly climatological information is available. The next critical task undertaken was to test whether the estimate of median LIS from MAP by regression equation for a specific Koppen climate class identified in southern Africa would remain similar for an identical Koppen climatic region in Eritrea. As already mentioned, LIS may be determined from daily time step hydrological soil moisture budget models such as ACRU model. The performance of the ACRU stormflow modelling approach was, therefore, first verified on an Eritrean gauged research catchment, viz. the Afdeyu, in order to have confidence in the use of values of LIS generated by it. A SCS-ACRU stormflow modelling approach was then tested on the same catchment by using the new approach of CN adjustment, termed the ACRU-Koppen method, and results were compared against stormflow volumes obtained using the SCS-AMC classes and the Hawkins' soil water budgeting procedures for CN adjustment, as well as when CNs remain unadjusted. Despite the relatively limited level of information on climate, soils and land use for the Afdeyu research catchment, the ACRU model simulated both daily and monthly flows well. By comparing the outputs generated from the SCS model when using the different methods of CN adjustment, the ACRU-Koppen method displayed better levels of performances than either of the other two SCS-based methods. A further statistical comparison was made among the ACRU, the SCS adjusted by ACRU-Koppen, the SCS adjusted by AMC classes and the unadjusted SCS models for the five highest stormflows produced from the five highest daily rainfall amounts of each year on the Afdeyu catchment. The ACRU model produced highly acceptable statistics from stormflow simulations on the Afdeyu catchment when compared to the SCS-based estimates. In comparing results from the ACRU-Koppen method to those from the SCS-AMC and unadjusted CN methods it was found that, statistically, the ACRU-Koppen performed much better than either of the other two SCS based methods. On the strength of these results the following conclusions were drawn: • Changes in soil moisture storage from so-called "initial" catchment soil moisture conditions, i.e. L1S, are similar in similar climatic regions; and • Using the ACRU-Koppen method ofCN adjustment, the SCS-SA model can, therefore, be adapted for application in Eritrea, for which Koppen climates can be produced from monthly rainfall and temperature maps. Finally, future research needs for improvements in the SCS-ACRU-Koppen (SAK) approach in light of data availability and the estimation ofL1S were identified. From the findings of this research and South African experiences, a first version of a "SCSEritrea" user manual based on the SAK modelling approach has been produced to facilitate its use throughout Eritrea. This user manual, although not an integral part of this dissertation, is presented in its entirety as an Appendix. A first Version of the SCS-Eritrea software is also included. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2004.

A combination of geological and hydraulic techniques represents the most sensible approach to flood hazard analysis on alluvial fans. Hydraulic models efficiently yield predictions of flood depths and velocities, but the assumptions on which the models are based do not lead to accurate portrayals of natural fan processes. Geomorphological mapping, facies, mapping, and hydraulic reconstructions of past floods provide data on the location, types, and magnitude of flood hazards, respectively. Geological reconstructions of past floods should be compared with the results of hydraulic modeling before, potentially unsound, floodplain management decisions are implemented. The controversial Federal Emergency Management Agency procedure for delineating flood-hazard zones underestimated the extent, velocity, and depth of flow during recent floods on two alluvial fans by over 100, 25, and 70 percent, respectively. Flow on the alluvial fans occurs in one or more discontinuous ephemeral stream systems characterized by alternating sheetflood zones and channelized reaches. The importance of sheetflooding is greater on fans closer to the mountain front and with unstable channel banks. Channel diversions on five alluvial fans repeatedly occurred along low channel banks and bends where the greatest amount of overland flow is generated. Channel migration occurs through stream capture whereby overland flow from the main channel accelerates and directs erosion of adjacent secondary channels. The recurrence interval of major channel shifts is greater than 100 years, but minor changes occurred on all five fans during this century. Small aggrading flows are important, because they decrease bank heights and alter the location of greatest overland flow during subsequent floods. The results of this study demonstrate that (1) geological reconstructions of past floods can check the results of hydraulic models, (2) the character of flooding on alluvial fans can vary significantly in the same tectonic and climatic setting due to differences in drainage-basin characteristics, and (3) flood-hazard assessments on alluvial fans must be updated after each flood, because the location and timing of channel diversions can be affected by small floods.

From the Proceedings of the 1971 Meetings of the Arizona Section - American Water Resources Assn. and the Hydrology Section - Arizona Academy of Science - April 22-23, 1971, Tempe, Arizona / The rotadisk rainulator is a recently developed rainfall simulator utilizing a full-cone-spray type nozzle. Its unique feature is the rotation of disks of various size openings that makes it possible to produce intensities from close to zero up to full nozzle capacity. Disks may be quickly changed, making it possible to study the effects of various intensities on infiltration rates, such as occur in natural storms. For all intensities above 1.0 in/hr, the instrument comes closer to duplicating kinetic energies and momenta of natural rainfall than any other type of rainfall simulator. Little rainfall-runoff data are available on most of the Lower Gila watersheds. Infiltration rates were therefore determined using the rotadisk rainulator on recompacted soil samples from the watershed. The results permitted a ranking of the watersheds on the basis of infiltration rates, which supports an independent flood frequency analysis indicating that the flood threat from subwatersheds along the Gila is much lower than had previously been projected. When the instrument is taken into the field, it should be possible to directly determine the infiltration rates of different soil and vegetation types, which will be of more use to hydrologists than data from recompacted samples