Paper/Poster/Presentation Title
DATA-DRIVEN RAINFALL/RUNOFF MODELLING BASED ON A NEURO-FUZZY INFERENTIAL SYSTEM
Keywords
Rainfall/runoff models; Principal Component Analysis; Neuro-Fuzzy Networks; ANFIS.
Location
Session C1: VI Data Mining for Environmental Sciences Session
Start Date
13-7-2016 10:50 AM
End Date
13-7-2016 11:10 AM
Abstract
The development of rainfall/runoff models may involve extensive computation and differing platforms, including GIS. In this paper we present a simple data-driven approach which avoids the use of GIS, but is based on a combination of Principal Component Analysis (PCA) and an Adaptive Neuro Fuzzy Inference System (ANFIS) to produce a simple and effective output flow prediction based on previous rainfall/runoff data. Given the ANFIS internal complexity, the emphasis of the paper is on how to set-up the most representative and parsimonious data structure that produces an efficient output flow estimation. In the preliminary data reduction stage, the PCA approach is compared to the equivalent rainfall computed by the Thiessen polygons, involving GIS, and it is demonstrated that the former approach yields a better data set for the ANFIS processing in terms of algorithm complexity and output accuracy. The algorithm is applied to two differing medium-size catchments in Tuscany, central Italy, and provides an excellent approximation of the output discharge with reduced computational complexity.
Included in
Civil Engineering Commons, Data Storage Systems Commons, Environmental Engineering Commons, Hydraulic Engineering Commons, Other Civil and Environmental Engineering Commons
DATA-DRIVEN RAINFALL/RUNOFF MODELLING BASED ON A NEURO-FUZZY INFERENTIAL SYSTEM
Session C1: VI Data Mining for Environmental Sciences Session
The development of rainfall/runoff models may involve extensive computation and differing platforms, including GIS. In this paper we present a simple data-driven approach which avoids the use of GIS, but is based on a combination of Principal Component Analysis (PCA) and an Adaptive Neuro Fuzzy Inference System (ANFIS) to produce a simple and effective output flow prediction based on previous rainfall/runoff data. Given the ANFIS internal complexity, the emphasis of the paper is on how to set-up the most representative and parsimonious data structure that produces an efficient output flow estimation. In the preliminary data reduction stage, the PCA approach is compared to the equivalent rainfall computed by the Thiessen polygons, involving GIS, and it is demonstrated that the former approach yields a better data set for the ANFIS processing in terms of algorithm complexity and output accuracy. The algorithm is applied to two differing medium-size catchments in Tuscany, central Italy, and provides an excellent approximation of the output discharge with reduced computational complexity.
