Premium
The Analysis of Structural Identifiability: Theory and Application to Conceptual Rainfall‐Runoff Models
Author(s) -
Sorooshian Soroosh,
Gupta Vijai Kumar
Publication year - 1985
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/wr021i004p00487
Subject(s) - identifiability , sensitivity (control systems) , identification (biology) , estimation theory , model selection , measure (data warehouse) , conceptual model , computer science , parameter space , model parameter , selection (genetic algorithm) , mathematical optimization , mathematics , econometrics , data mining , algorithm , statistics , machine learning , engineering , ecology , database , electronic engineering , biology
The identification of a conceptual hydrologic model involves two stages: selection of a suitable model structure and parameter estimation. These two stages cannot be treated as distinct, and it is important that the limitations of available parameter estimation methodologies be given serious consideration during the procedure of model structure selection. This paper focuses on the concepts and tools necessary for the evaluation of whether the structure of a hydrologic model is such that its parameters can be successfully identified. A definition of structural identifiability is advanced that is based purely on model properties and is independent of the stochastic nature of the output observation errors. The approach is used to develop a measure of local structural identifiability which is useful in the analysis of parameter sensitivity. A measure, called the sensitivity ratio, is introduced which is suitable for studying parameter interactions in the multidimensional parameter space. Also, the use of allowable reparameterizations to improve identifiability is discussed. The utility of the concepts discussed in this paper are demonstrated by a simulation example using a conceptual rainfall‐runoff model.