Improved Calibration Of Conceptual Hydrologic Models VIA GIS Aided Approach

The main requirements for an operational conceptual hydrologic model are function development, model calibration, and verification followed by production of forecast runs and analysis of results. By far, the calibration requirement may be the most critical stage of the overall modeling process. As easy as it may appear, the calibration of a conceptual hydrologic model (CHM) with the intention of identifying a 'unique ' and 'consistent' set of parameters is a tedious task, made complicated by a whole host of problems. In the past decade, many researchers have expressed concerns regarding the uniqueness of parameter estimates for CHMs obtained through calibration. It has been observed that even though stochastic parameter estimation techniques can help, the problems are not all due to the inefficiencies in the calibration techniques used but are caused by the manner in which the model is structurally formulated. Therefore it is essential to identify the problems cause by the model structure in the automatic calibration.The present study investigates the problems cause by the model structure in the automatic calibration of conceptual hydrologic models and the use of hydrological response units (HRUs) in CHMs to improve the model performance via integrated GIS approach while keeping the number of basic model parameters and the model structure as original. For the implementation of the concepts, the Xinanjiang hydrologic model has been used.Initially the Xinanjiang model was calibrated using the Shuffle Complex Evolution method (SCE-UA) under ideal situations -where the data inputs are assumed to be error free and the flow series for which the model results are fitted are generated from a known set of model parameter values. It was observed that the SCE-UA method could locate the global optimum parameter set with 100% success rate indicating that SCE-UA is capable of finding the global optimum parameter set in the automatic calibration of the Xinanjiang model.However it was unable to find a stable parameter set when using the real catchment data with large range of upper and lower boundary values. Hence it was identified that there were some problems caused by the model structure in the automatic calibration. The correlation analysis conducted on the Xinanjiang model parameters revealed that KG and KI, B and EX, B and CG, EX and CS, B and K are highly correlated. Though the highest correlation coefficient obtained between KG and KI, most of the problems encountered in the automatic calibration process are associated with the calibration of parameters B and EX. However further researchconducted on this matter indicated that most of the problems caused by the Xinanjiang model structure in the automatic calibration can be eliminated by fixing the parameters B and EX assigning appropriate values.The additional calibration trials conducted using historical data from the Misai, Wanjiabo and Kalu indicated that the SCE-UA algorithm is effective and efficient optimization tool for calibrating the Xinanjiang model and capable of finding a conceptually realistic and valid parameter set in the automatic calibration. All the Xinanjiang model parameters except B and EX can be calibrated simultaneously using the SCE-UA method. In addition the calibration trials conducted by incrementally adding the input data in yearly basis indicate that there exists a positive trend of convergence of the parameter values to a general or a unique parameter set for a particular catchment with the increment of the input data period.In the second phase of this study, the use of hydrological response units (HRUs) in conceptual hydrologic models (e.g., in the Xinanjiang model) is discussed and its application to the Kalu river upper catchment in Sri Lanka is presented. Instead of delineating the sub catchments based on the traditional Thiessen polygon method, the HRU concept has been implemented such that each HRU (based on the land-use classes) represents a sub catchment. The results were compared with the SWAT model (Soil and Wa...