A hybrid generalised linear and Levenberg–Marquardt artificial neural network approach for downscaling future rainfall in North Western England

Abstract

This paper describes a novel technique for downscaling daily rainfall which uses a combination of a Generalised Linear Model (GLM) and Artificial Neural Network (ANN) to downscale rainfall. A two-stage process is applied, an occurrence process which uses the GLM model and an amount process which uses an ANN model trained with a LevenbergMarquardt approach. The GLM-ANN was compared with other three downscaling models, the traditional neural network (ANN), multiple linear regression (MLR) and Possion regression (PR). The models are applied for downscaling daily rainfall at three locations in North West of England during the winter and summer. Model performances with respect to reproduction of various statistics such as correlation coefficient, autocorrelation, root mean square errors (RMSE), standard deviation and the mean rainfall are examined. It is found that the GLM-ANN model performs better than the other three models in reproducing most daily rainfall statistics, with slight difficulties in predicting extremes rainfall event in summer. The GLM-ANN model is then used to project future rainfall at the three locations employing three different GCMs for SRES scenarios A2 & B2. The study projects significant increases in mean daily rainfall at most locations for winter and decrease in summer.

Assessment of Water Harvesting System for a Smart Building Considering Climate Change

“…Future rainfall has been projected at the three locations in a previous study see [12] using a combination of Generalized Linear Model (GLM) and Artificial Neural Network ANN (hereinafter known as the hybrid GLM-ANN model, Fig. 2).…”

“…Unlike other downscaling techniques, the developed model uses two-stage process to model rainfall; an occurrence process which uses the GLM with a logistic regression model and an amount process which uses an ANN network model trained with Levenberg-Marquardt approach. The developed model was used to simulate future rainfall using climatic variable produced from HadCM3 GCM for emissions scenarios high (A1FI) and low (B1) scenarios to understand relationship between climate change and hydrology in the selected sites, see [12] for calibration, verification of this model. The following is the description of the occurrence model (GLM),…”

“…k = Number of predictors. A multi-layer feed forward artificial neural network (MLF-ANN) model was used to build a non-linear relation between the observed rainfall amount series and the same selected set of climatic variables (predictors) used for the rainfall occurrence model [12].…”

“…In order to verify whether the downscale rainfall model has the ability to capture the observed feature of each site an example of the verification has been presented below for both occurrence and amount (more details see [12]). …”

Monitoring the future behaviour of urban drainage system under climate change: a case study from north-western England

“…A hybrid Generalised Linear Model (GLM) and Artificial Neural Network (ANN) approach, developed in study [10], was used to downscale rainfall from coarse global climate model (GCM) (HadCM3) outputs to finer spatial scales. More on information on model and how it has been applied can be found in the given reference and will not be repeated here.…”