Predicting runoff and salinity intrusion using stochastic precipitation inputs
Committee ChairFogel, Martin M.
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractA methodology is presented for forecasting the probabilistic response of salinity movement in an estuary to seasonal rainfall and freshwater inflows. The Gambia River basin in West Africa is used as a case study in the research. The rainy season is from approximately July to October. Highest flows occur in late September and early October. Agriculturalists are interested in a forecast of the minimum distance that occurs each year at the conclusion of the wet season between the mouth of the river and the 1 part per thousand (ppt) salinity level. They are also interested in the approximate date that the minimum distance will occur. The forecasting procedure uses two approaches. The first uses a multisite stochastic process to generate long-term synthetic records (100 to 200 years) of 10-day rainfall for two stations in the upper basin. A long-term record of 10-day average flow is then computed from multiple regression models that use the generated rainfall records and real-time initial flow data occurring on the forecast date as inputs. The flow series is then entered into a one-dimensional finite element salt intrusion model to compute the movement of the 1 ppt salinity level for each season. The minimum distances between the mouth of the river and the 1 ppt salinity front that occurred for each season in the long-term record are represented in a cumulative probability distribution curve. The curve is then used to assign probability values of the occurrence of the 1 ppt salinity level to various points along the river. In the second approach, instead of generating a rainfall series and computing flow from regression models, a long-term flow record was generated using a stochastic first-order Markov process. Probability curves were made for three forecast dates: mid- July, mid-August, and mid-September using both approaches. With the first approach, the initial conditions at the time of the forecast had a greater influence on the flow series than the second approach.
Degree NamePh. D.
Degree ProgramRenewable Natural Resources