Runoff Prediction In Ungauged Catchments

Hydrological predictive tools have been developed from experimental equations,lumped conceptual models to distributed physical models.Most studies focuses on gauged catchments, that is,using the model with observed streamflow data.However,there are still a lot ofungaged catchments in the world and some gauged catchments are becoming ungauged because of the impact of environmental change.So major problems are encountered when applying these models to ungaged catchments for which no streamflow data are available and hence no such calibration is possible.This problem is getting worse with the climate change and land use change,which makes runoff prediction more uncertain and complex.In this paper,210 relatively unimpaired gauged catchments(50 to 2000 km~2) in south-east Australia are taken as ungauged ones for runoff prediction in ungauged catchments.First,a better regionalization method is proposed.Then,in order to improve runoff prediction in ungauged catchments,on one hand,an improved Penman-Monteith equation which include remotely sensed leaf area index and the Xinanjiang model are used for catchment evapotranspiration prediction and rainfall-runoff simulation;on the other hand,multi-objective calibration is used to reduce the prediction uncertainty.The main research results are summeried as follows:(1) Three classical regionalization schemes of catchment model parameters over the 210 unimpaired catchments in south-east Australia:spatial proximity,physical similarity and regression are compared.The results show that in Australia,the spatial proximity approach provides the best regionalization solution,the physical similarity approach is intermediary,the regression approach is the least satisfactory.But there exists some complementary between the spatial proximity approach and the physical similarity approach.So a new method,integrated similarity,is proposed by combining the spatial proximity and physical similarity.In integrated similarity approach,distance between catchment centroids together with the other catchment attributes are used as similarity descriptors to describe similarity of a donor(gauged) catchment to the ungauged catchment,and then the complete parameter set from each of the donor catchments is transferred to the ungauged catchment.The results show that the integrated similarity approach further improved runoff prediction in ungauged catchments.In addition,a classification-regression method is proposed based on the regression method.In the classification-regression method,catchments are devided into three classes based on the precipitation,and then the regression method is used in each class.The results show that the classification-regression method is better than the regression method in predicting runoff in ungauged catchments.(2) Accurate estimates of evapotranspiration are required to reduce uncertain in constructing water balance in catchment and regional scales.Water used by plants,and lost by soil evaporation,can strongly influence the magnitude and variability of catchment water yield. For this reason,improved estimates of evapotranspiration(transpiration,soil evaporation) should improve estimates of catchment water yield,especially in ungauged catchments.We introduce a simple biophysical model for use remotely sensed leaf area index data and the Penman-Monteith equation to calculate daily-average evapotranspiration,based on understanding of radiation absorption by plant canopies and evaporation from the underlying soil surface.Modelled daily evapotranspiration compared well with measurents,indicated by a R~2 of 0.80 and the Nash-Sutcliffe coefficient of efficiency of 0.77.The transpiration is about 70%of the whole evapotranspiration,which is much stronger than the soil evaporation,and this result is similar to most other studies.This study confirms that the Penman-Monteith equation with remotely sensed leaf area index provides reliable estimates of evapotranspiration at daily time scales and regional scales.(3) Vegetation processes are seldom considered in lumped conceptual rainfall-runoff models although they have significant impacts on runoff via the control of evapotranspiration. Vegetation modulates surface-subsurface interactions via evapotranspiration which affects both the surface energy balance processes and water storage availability in the root-zone,and therefore impacts rainfall-runoff response.This paper incorporates the remotely sensed MODIS-LAI(the MODerate resolution Imaging Spectrometer mounted on the polar-orbiting Terra satellite - Leaf Area Index) data into Xinanjiang rainfall-nmoffmodel and assesses the model performance on 210 catchments in south-east Australia.The results show that the inclusion of LAI data improves both the model calibration results as well as the dally runoff prediction in ungauged catchments.(4) Multiple objectives calibration of Xinanjiang rainfall-runoff model to minimize the uncertainty in the modeled hydrological response is presented.The multi-objective calibration includes:first,optimization of multiple objectives that measures different aspects of the hydrograph:(a) peak flow and(b) low flow,and second,optimization of multiple objectives that measures different data of the hydrological field:(a) streamflow and(b) evapotranspiration. Significant trade-offs between different objectives are observed in this case and no single unique set of parameter values is able to optimise all objectives simultaneously.Multi-objective calibration reduced the uncertainty of runoff prediction in ungauged catchments.The benefits of multi-objective calibration are also illustrated when compared with traditional single objective calibrations.The results show that the single-objective calibration matches one aspect of the hydrograph well but is inadequate in the other aspect;the multi-objective calibration provides a proper balance between the two objectives and matches both aspects well.