Yuk Feng Huang

Application of the generalized likelihood uncertainty estimation (GLUE) approach for assessing uncertainty in hydrological models: a review

The generalized likelihood uncertainty estimation (GLUE) technique is an innovative uncertainty method that is often employed with environmental simulation models. Over the past years, hydrological literature has seen a large increase in the number of papers dealing with uncertainty. There are now a lot of citations to their original paper which illustrates GLUE tremendous impact. GLUE’s popularity can be attributed to its simplicity and its applicability to nonlinear systems, including those for which a unique calibration is not apparent. The GLUE was introduced for use in uncertainty analysis of watershed models has now been extended well beyond rainfall-runoff watershed models. Given the widespread adoption of GLUE analyses for a broad range or problems, it is appropriate that the validity of the approach be examined with care. In this article, we present an overview of the application of GLUE for assessing uncertainty distribution in hydrological models particularly surface and subsurface hydrology and briefly describe algorithms for sampling of the prior parameter in hydrologic simulation models.

Uncertainty analysis for extreme flood events in a semi-arid region

Extreme flood events are complex and inherently uncertain phenomenons. Consequently forecasts of floods are inherently uncertain in nature due to various sources of uncertainty including model uncertainty, input uncertainty, and parameter uncertainty. This paper investigates two types of natural and model uncertainties in extreme rainfall–runoff events in a semi-arid region. Natural uncertainty is incorporated in the distribution function of the series of annual maximum daily rainfall, and model uncertainty is an epistemic uncertainty source. The kinematic runoff and erosion model was used for rainfall–runoff simulation. The model calibration scheme is carried out under the generalized likelihood uncertainty estimation framework to quantify uncertainty in the rainfall–runoff modeling process. Uncertainties of the rainfall depths—associated with depth duration frequency curves—were estimated with the bootstrap sampling method and described by a normal probability density function. These uncertainties are presented in the rainfall–runoff modeling for investigation of uncertainty effects on extreme hydrological events discharge and can be embedded into guidelines for risk-based design and management of urban water infrastructure.

Stochastic modelling of seasonal and yearly rainfalls with low-frequency variability

Stochastic rainfall models are important for many hydrological applications due to their appealing ability to simulate synthetic series that resemble the statistical characteristics of the observed series for a location of interest. However, an important limitation of stochastic rainfall models is their inability to preserve the low-frequency variability of rainfall. Accordingly, this study presents a simple yet efficient stochastic rainfall model for a tropical area that attempts to incorporate seasonal and inter-annual variabilities in simulations. The performance of the proposed stochastic rainfall model, the tropical climate rainfall generator (TCRG), was compared with a stochastic multivariable weather generator (MV-WG) in various aspects. Both models were applied on 17 rainfall stations at the Kelantan River Basin, Malaysia, with tropical climate. The validations were carried out on seasonal (monsoon and inter-monsoon) and annual basis. The third-order Markov chain of the TCRG was found to perform better in simulating the rainfall occurrence and preserving the low-frequency variability of the wet spells. The log-normal distribution of the TCRG was consistently better in modelling the rainfall amounts. Both models tend to underestimate the skewness and kurtosis coefficient of the rainfall. The spectral correction approach adopted in the TCRG successfully preserved the seasonal and inter-annual variabilities of rainfall amounts, whereas the MV-WG tends to underestimate the variability bias of rainfall amounts. Overall, the TCRG performed reasonably well in the Kelantan River Basin, as it can represent the key statistics of rainfall occurrence and amounts successfully, as well as the low-frequency variability.

Temporal precipitation trend analysis at the Langat River Basin, Selangor, Malaysia

The Langat River Basin provides fresh water for about 1.2 million people in the Langat and Klang valleys. Any change in the pattern of rainfall could affect the quantity of water in the basin. Studying the pattern of change in rainfall is crucial for managing the available water resources in the basin. Thus, in this study, for the first time, both parametric and non-parametric methods were employed to detect rainfall trend in the basin for the period 1982–2011. The trends were determined at 30 rainfall stations using the Mann–Kendall (MK) test, the Sen’s slope estimator and the linear regression analysis. Lag-1 approach was utilized to test the serial correlation of the series. On the annual scale, it was found that most of the stations in the basin were characterized with insignificant trends. The significant trends were found only at the four stations, namely 44301, 44305, 44320 and 2719001. The results of the seasonal trend analysis showed that most of the stations during the northeast monsoon (NEM) and the inter monsoon 1 (INT1) seasons and half of the stations during the southwest monsoon (SWM) season experienced insignificant positive trends. To the contrary, for the inter monsoon 2 (INT2) season, majority of the stations showed negative trends. It was found that during the NEM season the station 44301, for the INT1 season stations 44301, 2719001 and 3118069 were established as having significant changes, while in the SWM season station 2917001 and during the INT2 season, the stations 2615131 and 44301 showed significant trends. It is worth mentioning that the maximum rainfall occurs in inter-monsoon seasons.

Generation of a stochastic precipitation model for the tropical climate

A tropical country like Malaysia is characterized by intense localized precipitation with temperatures remaining relatively constant throughout the year. A stochastic modeling of precipitation in the flood-prone Kelantan River Basin is particularly challenging due to the high intermittency of precipitation events of the northeast monsoons. There is an urgent need to have long series of precipitation in modeling the hydrological responses. A single-site stochastic precipitation model that includes precipitation occurrence and an intensity model was developed, calibrated, and validated for the Kelantan River Basin. The simulation process was carried out separately for each station without considering the spatial correlation of precipitation. The Markov chains up to the fifth-order and six distributions were considered. The daily precipitation data of 17 rainfall stations for the study period of 1954–2013 were selected. The results suggested that second- and third-order Markov chains were suitable for simulating monthly and yearly precipitation occurrences, respectively. The fifth-order Markov chain resulted in overestimation of precipitation occurrences. For the mean, distribution, and standard deviation of precipitation amounts, the exponential, gamma, log-normal, skew normal, mixed exponential, and generalized Pareto distributions performed superiorly. However, for the extremes of precipitation, the exponential and log-normal distributions were better while the skew normal and generalized Pareto distributions tend to show underestimations. The log-normal distribution was chosen as the best distribution to simulate precipitation amounts. Overall, the stochastic precipitation model developed is considered a convenient tool to simulate the characteristics of precipitation in the Kelantan River Basin.

Downscaling daily precipitation and temperatures over the Langat River Basin in Malaysia: A comparison of two statistical downscaling approaches

Increasing greenhouse gas concentrations can cause future changes in the climate system that have a major impact on the hydrologic cycle. To realize and predict future climate parameters, the Atmosphere-Ocean Global Climate Models (AOGCMs) are common employed tools to predict the future changes in climate parameters. The statistical downscaling methods have been applied as a practical tool to bridge the spatial difference between grid-box scale and sub-grid box scale. This paper investigates the capability of Statistical Downscaling Model (SDSM) and Artificial Neural Network (ANN) with different complexities in downscaling and projecting climate variables in the tropical Langat River Basin. These two statistical downscaling models have been calibrated, validated and used to project the possible future scenarios (2030s and 2080s) of meteorological variables, which are the maximum and minimum temperatures as well as precipitation using the CGCM3.1 under A2 emission scenario. The statistical validation of generated precipitation as well as maximum and minimum temperatures on a daily scale illustrated that the SDSM is more accurate than the ANN with different learning rules. On the other hand, the SDSM showed more capability to catch the wet-spell and dry-spell lengths than the ANN model. The calibrated models show higher accuracy in simulating the maximum and minimum temperatures in comparison with the capture of the variability of precipitation. The trend analysis test of generated time series by the SDSM indicates an increasing trend by the 2030s and 2080s at most of the stations. Key words: Statistical downscaling, multiple linear regression, nonlinear regression, artificial neural network, tropical area, Malaysia.

River catchment rainfall series analysis using additive Holt–Winters method

Climate change is receiving more attention from researchers as the frequency of occurrence of severe natural disasters is getting higher. Tropical countries like Malaysia have no distinct four seasons; rainfall has become the popular parameter to assess climate change. Conventional ways that determine rainfall trends can only provide a general result in single direction for the whole study period. In this study, rainfall series were modelled using additive Holt–Winters method to examine the rainfall pattern in Langat River Basin, Malaysia. Nine homogeneous series of more than 25 years data and less than 10% missing data were selected. Goodness of fit of the forecasted models was measured. It was found that seasonal rainfall model forecasts are generally better than the monthly rainfall model forecasts. Three stations in the western region exhibited increasing trend. Rainfall in southern region showed fluctuation. Increasing trends were discovered at stations in the south-eastern region except the seasonal analysis at station 45253. Decreasing trend was found at station 2818110 in the east, while increasing trend was shown at station 44320 that represents the north-eastern region. The accuracies of both rainfall model forecasts were tested using the recorded data of years 2010–2012. Most of the forecasts are acceptable.