The Monthly And Seasonal Streamtlow Prediction Using Parameterization From Data Of Diffcrent Calibration Periods

Reliable seasonal predictions of streamflows are highly valuable and will have been used in providing water allocation outlooks, informing water markets, planning and managing water use, managing flood and drought and national defense construction. In terms of its climate, Australia has experienced a remarkable decade. The continent has had its warmest period since records began and southern areas have been extremely dry. A monthly and seasonal streamflow prediction service has been needed in Australia for many years and the Australian Government's recent investment in water information will help address this need. A seasonal climate prediction service has been operating in the Bureau of Meteorology since1980s but its primary focus has been on rainfall and temperature rather than water availability. The medium to long-term dynamical ensemble streamflow forecasting (ESP) based on the hydrologic model and numerical weather prediction become more and more important, although the statistics and artificial intelligence approaches of less theoretical and universal has been widely used in the seasonal streamflow prediction. In the forecasting mode of ESP, a hydrologic model calibrated using historical data will be run forward, together with forecasted or re-sampled (from history) meteorological forcings, to predict the streamflows in the coming month or season. Three major factors control the forecasting accuracy:1) the performance of the hydrologic model to predict streamflow with actual forcings,2) the accuracy of the model-simulated catchment initial conditions (soil moisture and groundwater stores) at the forecasting time, and3) the accuracy of the GCM forecasts for rainfall and other related variables, or the accuracy of re-sampled meteorological forcings to represent the future forcings. How well the model can simulate the catchment conditions (soil moisture and groundwater stores) is also dependent on the performance of the model to simulate streamflow with actual forcings. Therefore, it is essential to carefully design model calibration procedures in order to ensure a good performance of the hydrologic model for streamflow forecasting. This dissertation investigate how the selection of model calibration periods, i.e., different parameterization schemes, can improve performance of a rainfall-runoff models, SIMHYD and Xinanjiang, for streamflow simulation in the coming year and quantify the skills of the rainfall-runoff model-based forecasts of streamflow at monthly and seasonal scales using both historical ensemble forcings and forcings derived from POAMA downscaling, and investigate the benefits of an alternative model calibration approach conditional to streamflows in each individual month in terms of model performance and forecasting skills for12catchments in east Australia. The main contents and results are as follows:Based on the mechanism of the rainfall-runoff model and hydrologic cycle, the change of the annual and monthly of precipitation, PET and runoff; and the relationship between the annual and monthly precipitation and runoff; ENSO event and atmospheric circulation factor which effect on the hydrologic situation (precipitation and runoff) are analyzed in the paper. It gives a material and theoretical support for the study of the medium and long-term hydrological forecasting based on the parameterization of rainfall-runoff model.Better parameterization of hydrological model can lead to improved streamflow prediction. This is particularly important for seasonal streamflow forecasting using hydrological modeling. Considering the possible effects of hydrologic non-stationarity, this dissertation examined10parameterization schemes at twelve catchments, located in three different climatic zones in east Australia. These schemes are grouped into four categories based on the period where the data are used for model calibration, i.e., calibration using:1) data from a fixed period in the historical records,2) different lengths of historical records prior to prediction year,3) data from different climatic analogue years in the past,4) data from individual month or season. Parameterization schemes were evaluated according to model efficiency in both calibration and verification period. The results show that the calibration skill changes with different historic periods where data are used at all catchments. Comparison of model performance shows that the parameterization schemes of SIMHYD is better than Xinanjiang rainfall-runoff model.Comparison of model performance between calibration schemes indicates that it is worth calibrating the model using data from each individual month for purpose of seasonal streamflow forecasting. For the catchments in winter-dominant rainfall region of southeast Australia, more significant shift in rainfall-runoff relationships in different periods was found. For those catchments, model calibration using20years of data prior to the prediction year leads to more consistent perforemance.Well-validated rainfall-runoff models are able to capture the relationships between rainfall and streamflow and to reliably estimate initial catchment states. While future streamflows are mainly dependent on initial catchment states and future rainfall, use of the rainfall-runoff model together with estimated future rainfall can produce skilful forecasts of future streamflows. This is the basis for the ensemble streamflow prediction system, but this approach has not been explored in Australia. This dissertation explored the skills of forecasts for monthly and three-monthly total streamflows in east Australia with a dynamic approach using a conceptual rainfall-runoff model SIMHYD. Two model calibration approaches were applied; both used the data before the forecast year to calibrate the SIMHYD model in a moving mode. The first approach, i.e., general parameterisation, used all the data from calibration period to calibrate the model, leading to one set of parameters (S7). The second approach, called conditional parameterisation, only used data from the target forecast month/season prior to the forecast year for model calibration, leading to12sets of parameters, each for a given month/season (S10). In addition, two types of forcing were used to represent future rainfall and potential evapotranspiration, one re-sampled from the whole historical data, another from the ensemble forecasts of the Predictive Ocean Atmosphere Model for Australia (POAMA).It was found that the dynamic forecasting approach based on conceptual rainfall-runoff modelling provides a potential way to improve streamflow forecasting at monthly and seasonal lead time in east Australia. The conditional parameterisation approach improved model performance in both calibration and verification period, and more significantly the skills of monthly and three-monthly forecasts with both historical and POAMA ensemble forcings. In this dissertation, the conceptual rainfall-runoff model SIMHYD and two paramterization schemes (S7and S10), together with rainfall and PET ensembles based on historical rainfall and PET, were used to forecast streamflows at monthly and3-monthly scales at12catchments in east Australia. The results of ensemble streamflow prediction showed that SIMHYD rainfall-runoff model forecast monthly and3-monthly streamflow well when forecasts for all months and seasons were evaluated together, but their performance varied significantly from month (season) to month (season). Best forecasting skills were obtained (both monthly and3monthly) when the models were coupled with ensemble forcings on the basis of long-term historical rainfall. The paramterization scheme S10of SIMHYD and forcings from historical data led to improved forecasting skills. For the catchments in winter-dominant rainfall region of southeast Australia, the rainfall-runoff model showed more better and consistent perforemance in monthly and3monthly streamflow prediction.Using POAMA forecasts as forcing for the rainfall-runoff model improved the monthly streamflow forecasting skills as compared to using historical ensemble, and poor in3monthly streamflow forecasting performance. Combining POAMA forecasts and conditional parameterization led to better forecasts than climatology in65.3%of the months across the12 catchments for monthly forecasts, and43.1%of the months for three-monthly forecasts. Comparing the catchments in winter-dominant rainfall region to others, the conditional parameterization led to more better and consistent perforemance for monthly and3monthly streamflow prediction.