Yen-Ming Chiang

Multi-step-ahead neural networks for flood forecasting

Abstract A reliable flood warning system depends on efficient and accurate forecasting technology. A systematic investigation of three common types of artificial neural networks (ANNs) for multi-step-ahead (MSA) flood forecasting is presented. The operating mechanisms and principles of the three types of MSA neural networks are explored: multi-input multi-output (MIMO), multi-input single-output (MISO) and serial-propagated structure. The most commonly used multi-layer feed-forward networks with conjugate gradient algorithm are adopted for application. Rainfall—runoff data sets from two watersheds in Taiwan are used separately to investigate the effectiveness and stability of the neural networks for MSA flood forecasting. The results indicate consistently that, even though the MIMO is the most common architecture presented in ANNs, it is less accurate because its multi-objectives (predicted many time steps) must be optimized simultaneously. Both MISO and serial-propagated neural networks are capable of performing accurate short-term (one- or two-step-ahead) forecasting. For long-term (more than two steps) forecasts, only the serial-propagated neural network could provide satisfactory results in both watersheds. The results suggest that the serial-propagated structure can help in improving the accuracy of MSA flood forecasts.

Satellite‐based precipitation estimation using watershed segmentation and growing hierarchical self‐organizing map

This paper outlines the development of a multi‐satellite precipitation estimation methodology that draws on techniques from machine learning and morphology to produce high‐resolution, short‐duration rainfall estimates in an automated fashion. First, cloud systems are identified from geostationary infrared imagery using morphology based watershed segmentation algorithm. Second, a novel pattern recognition technique, growing hierarchical self‐organizing map (GHSOM), is used to classify clouds into a number of clusters with hierarchical architecture. Finally, each cloud cluster is associated with co‐registered passive microwave rainfall observations through a cumulative histogram matching approach. The network was initially trained using remotely sensed geostationary infrared satellite imagery and hourly ground‐radar data in lieu of a dense constellation of polar‐orbiting spacecraft such as the proposed global precipitation measurement (GPM) mission. Ground‐radar and gauge rainfall measurements were used to evaluate this technique for both warm (June 2004) and cold seasons (December 2004–February 2005) at various temporal (daily and monthly) and spatial (0.04° and 0.25°) scales. Significant improvements of estimation accuracy are found classifying the clouds into hierarchical sub‐layers rather than a single layer. Furthermore, 2‐year (2003–2004) satellite rainfall estimates generated by the current algorithm were compared with gauge‐corrected Stage IV radar rainfall at various time scales over continental United States. This study demonstrates the usefulness of the watershed segmentation and the GHSOM in satellite‐based rainfall estimations.

A spatial neural fuzzy network for estimating pan evaporation at ungauged sites

Evaporation is an essential reference to the management of water resources. In this study, a hybrid model that integrates a spatial neural fuzzy network with the kringing method is developed to estimate pan evaporation at ungauged sites. The adaptive network-based fuzzy inference system (ANFIS) can extract the nonlinear relationship of observations, while kriging is an excellent geostatistical interpolator. Three-year daily data collected from nineteen meteorological stations covering the whole of Taiwan are used to train and test the constructed model. The pan evaporation (Epan) at ungauged sites can be obtained through summing up the outputs of the spatially weighted ANFIS and the residuals adjusted by kriging. Results indicate that the proposed AK model (hybriding ANFIS and kriging) can effectively improve the accuracy of Epan estimation as compared with that of empirical formula. This hybrid model demonstrates its reliability in estimating the spatial distribution of Epan and consequently provides precise Epan estimation by taking geographical features into consideration.

A hybrid artificial neural network-based agri-economic model for predicting typhoon-induced losses

Building a model to rapidly simulate the impact of typhoons on agriculture and to predict agricultural losses is crucial and great help for remedial measure and distributing subvention right after the disaster. The relationship between typhoon-related meteorological factors and agricultural losses was first evaluated, and the Pearson’s test was applied to find consequences of both landfall and non-landfall which can be appropriately used to synthesize the possible coverage to suitably describe how typhoons influence agricultural losses. The self-organizing feature map (SOM) was then used to map similar properties of data into the same cluster and display the distribution of input–output patterns. Then, the clusters were adopted as centroids of radial basis function (RBF) neural networks. Finally, two hybrid self-organizing radial basis (SORB) networks that integrated SOM into RBF were constructed for predicting the event-based agricultural losses by feeding two different meteorological inputs (scenarios 1 and 2). The results indicate that the constructed SORB network has great ability to capture the relationship between meteorological characteristics and agricultural losses. Previously, it always takes several days to investigate and evaluate the agricultural damages after typhoons, which is a time-consuming process. In this study, the proposed agri-economic model also demonstrates its outstanding predictability, in real-time, and therefore effectively accelerates the official decision making on agricultural compensation after a typhoon strike.

Self-organizing radial basis neural network for predicting typhoon-induced losses to rice

The issue of the typhoon-induced economic losses to rice is investigated. In this study, we propose a hybrid self-organizing radial basis (SORB) neural network for estimating economic losses of rice for the whole Taiwan as well as three sub-regions. The data sets of 143 typhoon events from 1961 to 2008 were collected and analyzed. Data include rice losses and typhoon-related meteorological factors. A number of different input combinations of meteorological and temporal variables are implemented to select the optimal network for predicting the losses, and a two-stage clustering method is used to explore the spatial classification of 15 counties in Taiwan into three sub-regions. The simulation results indicate that the constructed SORB network has a great ability to capture the relationship between typhoon-related variables and rice losses. Furthermore, the SORB model also demonstrates its outstanding reliability and predictability for efficiently providing a valuable reference for counties in Taiwan that could protect farmers from exposure to increasing weather-related risk and accelerate the official decision making process on compensation for rice losses after the invasion of typhoons.

Evaluating the contribution of multi-model combination to streamflow hindcasting by empirical and conceptual models

ABSTRACT The contribution of multi-model combination to daily streamflow hindcasting was evaluated through the HBV (Hydrologiska Byråns Vattenbalansavdelning) and RNN (recurrent neural networks) models with 100 ensemble members generated with different initial conditions for both. In the calibration phase, the analysis showed that the HBV and RNN models with 20 members have better accuracy and require less calibration time. The combination of two models, however, did not provide significant improvements when 80 more members were added in the combination. In the validation phase, the results indicated that both HBV and RNN models with 20 members not only accurately produce reliable and stable streamflow hindcasting, but also effectively simulate the timing and the value of peak flows. From the consistency of calibration and validation results, the study provides an important contribution, namely, that ensemble size is not sensitive to the type of hydrological model in terms of streamflow hindcasting.

A two-step sensitivity analysis for hydrological signatures in Jinhua River Basin, East China

ABSTRACT Parameter calibration and sensitivity analysis (SA) are usually not straightforward tasks for distributed hydrological models, owing to the complexity of models and the large number of parameters. A two-step SA approach is proposed for analysing hydrological signatures based on the distributed hydrology–soil–vegetation model (DHSVM) in the Jinhua River Basin, East China. A preliminary SA is conducted to obtain influential parameters via analysis of variance. These parameters are further analysed through a variance-based global sensitivity analysis method to achieve robust rankings and parameter contributions. Parallel computing is designed to reduce the computational burden. The results reveal that only a few parameters are significantly sensitive and that interactions between parameters cannot be ignored. When analysing hydrological signatures, it is found that water yield is simulated very well for most samples. Small and medium floods are simulated very well, while slight underestimations happen for large floods.