Using recurrent neural networks to reconstruct rainfall‐runoff processes

Abstract

Abstract:Many novel techniques for reconstructing rainfall-runoff processes require hydrometeorologic and geomorphologic information for modelling. However, certain information is not always measurable. In this paper, we employ a special recurrent neural network to reconstruct the rainfall-runoff process by using collected rainfall data. In addition, we propose an indirect system identification to overcome the drawback of a traditional, time-consuming trial-and-error search. The indirect system identification is an efficient method to recognize the structure of a recurrent neural network. The unit hydrograph can be derived directly from the weights of the network due to a state-space form embedded in the recurrent neural network. This improves the link between the weights of the network and the physical concepts that most neural networks fail to connect.The case studies of 41 events from 1966 to 1997 have been implemented in Taiwan's Wu-Tu watershed, where the runoff path-lines are short and steep. Two recurrent neural networks and one state-space model are utilized to simulate the rainfall-runoff processes for comparison. The results are validated by four criteria: coefficient of efficiency; peak discharge error; time to peak arrival error; total discharge volume error. The resulting data from the recurrent neural network reveal that the neural network proposed herein is appropriate for hydrological systems.

Radial basis function neural network for hydrologic inversion: an appraisal with classical and spatio-temporal geostatistical techniques in the context of site characterization

“…Furthermore, most applications of ANNs in hydrology appear to rely heavily on the Back Propagation (BP-ANN) as the most popular algorithm (ASCE Task Committee 2000a; Maier and Dandy 2000). Examples of applications of BP-ANN are in rainfall-runoff modeling (Antar et al 2006;Pan and Wang 2005), in openchannel flow modeling (Lu et al 2006;Ahmad and Simonovic 2005;Chang et al 2002;Sivakumar et al 2002), in flood forecasting (Chen et al 2005;Huang et al 2004), in sediment transport modeling (Cigizoglu and Kisi 2006), and in groundwater applications (Garcia and Shigidi 2006;Akin 2005;Khalil et al 2005;Almasri and Kaluarachchi 2005;Lallahem et al 2005;Mantoglou 2003). Fortunately, some works do indicate that a collective effort is growing to gain understanding of ANN applications in hydrology based on the physical process it models.…”

Two decades of anarchy? Emerging themes and outstanding challenges for neural network river forecasting

“…Some papers have compared FFNN, Radial Basis Function NN (RBNN) and General Regression NN (GRNN) (Kişi, 2008a; Kişi and Cigizoglu, 2007), or Bayesian NN (FFNN calibrated using the routines provided in Nabney, 2002) with a standard FFNN and a conceptual rainfall-runoff model (Khan and Coulibaly, 2006). Others favoured dynamic models with simple feedback loops using a Partial-Recurrent Neural Network (PRNN: Besaw et al, 2010; Carcano et al, 2006, 2008; Chang et al, 2002, 2004; Chiang et al, 2004; Coulibaly et al, 2001a; Kumar et al, 2004; Pan and Wang, 2005). This refinement can have a profound impact since the model’s internal state depends not only on the current input signal, but also on its previous condition.…”

Improvement of watershed flood forecasting by typhoon rainfall climate model with an ANN-based southwest monsoon rainfall enhancement

“…Through the indirect system identification algorithm proposed by Pan and Wang (2005), the SSNN of the Tsengwen Reservoir watershed is built as a single-input-single-output model with 5-order state space based on the effective observed rainfall and direct runoff. Table 3 gives the performances of the hydrological model based on different estimated rainfall for 1-h-ahead flood forecasting of six typhoons.…”