Clustering of heterogeneous precipitation fields for the assessment and possible improvement of lumped neural network models for streamflow forecasts

Abstract

Abstract. This work addresses the issue of better considering the heterogeneity of precipitation fields within lumped rainfall-runoff models where only areal mean precipitation is usually used as an input. A method using a Kohonen neural network is proposed for the clustering of precipitation fields. The evaluation and improvement of the performance of a lumped rainfall-runoff model for one-day ahead predictions is then established based on this clustering. Multilayer perceptron neural networks are employed as lumped rainfallrunoff models. The Bas-en-Basset watershed in France, which is equipped with 23 rain gauges with data for a 21-year period, is employed as the application case. The results demonstrate the relevance of the proposed clustering method, which produces groups of precipitation fields that are in agreement with the global climatological features affecting the region, as well as with the topographic constraints of the watershed (i.e., orography). The strengths and weaknesses of the rainfall-runoff models are highlighted by the analysis of their performance vis-à-vis the clustering of precipitation fields. The results also show the capability of multilayer perceptron neural networks to account for the heterogeneity of precipitation, even when built as lumped rainfallrunoff models.

Including spatial distribution in a data-driven rainfall-runoff model to improve reservoir inflow forecasting in Taiwan

“…() and Lauzon et al . (), which are modified by the adoption of physically meaningful sub‐catchment boundaries, developed using drainage network analysis of a DEM. To this extent, a basic level of meaningful, hydrological process knowledge is incorporated into the data‐driven modelling operation.…”

“…The reported method represents a significant advance over the previous use of lumped mean areal inputs (Lorrai and Sechi, 1995), or distributed point-based rainfall sampling sites (Campolo et al, 1999;Dawson et al, 2006), given that such past approaches possess no explicit physical or operational underpinnings. It is also a physical and structural enrichment of previous data-driven radar rainfall modelling procedures applied by Teschl and Randeu (2006), and of the spatial clustering arguments of Rajurkar et al (2002) and Lauzon et al (2006), which are modified by the adoption of physically meaningful sub-catchment boundaries, developed using drainage network analysis of a DEM. To this extent, a basic level of meaningful, hydrological process knowledge is incorporated into the data-driven modelling operation.…”

Diagnostic evaluation of conceptual rainfall–runoff models using temporal clustering

“…The concept has been applied in hydrology to cluster, e.g. precipitation fields (Lauzon et al , 2006), watershed conditions (Liong et al , 2000), hydrological homogeneous regions (Frapporti et al , 1993; Hall and Minns, 1999), and also for regionalization purposes (Burn, 1989; Srinivas et al , 2008) and for hydrological model evaluation and identification (Herbst et al , 2009).…”

Non-linear visualization and analysis of large water quality data sets: a model-free basis for efficient monitoring and risk assessment

“…In some recent studies, SOMs were used to cluster discharge data for optimizing data selection for training of empirical runoff models (Liong et al 2000;Abrahart and See 2000;Hsu et al 2002;Jain and Srinivasulu 2006). Following a similar approach, Lauzon et al (2006) classified precipitation fields, Lauzon et al (2004) soil moisture profiles, and Lin and Chen (2005) groundwater head data as input into runoff models. Bowden et al (2004) clustered water quality data in order to optimize the selection of training data for a multilayer perceptron neural network.…”