Elsevier

Journal of Hydrology

Volume 517, 19 September 2014, Pages 1145-1153
Journal of Hydrology

Prediction of design flood discharge by statistical downscaling and General Circulation Models

https://doi.org/10.1016/j.jhydrol.2014.06.028Get rights and content

Highlights

  • We develop a model for predicting monthly peak discharge.

  • Hydrometeorological data is driven by downscaling.

  • NCEP and CGCM weather data is modeled by LGP.

  • Future climate scenarios are developed.

  • Most effective parameters are proposed for PMF prediction.

Summary

The global warming and the climate change have caused an observed change in the hydrological data; therefore, forecasters need re-calculated scenarios in many situations. Downscaling, which is reduction of time and space dimensions in climate models, will most probably be the future of climate change research. However, it may not be possible to redesign an existing dam but at least precaution parameters can be taken for the worse scenarios of flood in the downstream of the dam location. The purpose of this study is to develop a new approach for predicting the peak monthly discharges from statistical downscaling using linear genetic programming (LGP). Attempts were made to evaluate the impacts of the global warming and climate change on determining of the flood discharge by considering different scenarios of General Circulation Models. Reasonable results were achieved in downscaling the peak monthly discharges directly from daily surface weather variables (NCEP and CGCM3) without involving any rainfall–runoff models.

Introduction

For a hydraulic structure planned within the river, like a dam or a barrage, due consideration should be given to the design of the structure so as to prevent it from collapsing and causing further damage by the force of water released from behind the structure. Hence, an estimate of design flood is required. The magnitude of such flood may be estimated in accordance with the importance of the structure. The Probable Maximum Flood (PMF) is the extreme flood that may be expected from the most severe combination of critical meteorological and hydrologic conditions that are reasonably possible in a particular drainage area. In some situations where substantial risk for loss of life exists, it is appropriate to design the hydraulic structure against the worst possible condition, which is the PMF. The PMF is determined based on the unit hydrograph and probable maximum precipitation. In general, PMF is accepted as the standard for the design of spillway of dams where failure could lead to catastrophic situation including loss of life (Usul, 2005). Meanwhile when PMF is not applicable or possible, Flood Frequency Analysis (FFA) method used to determine the design flood. The objective of this method is to relate the magnitude of events to their frequency of occurrence through theoretical probability distributions based on determined return period. For instance, spillways for major and medium projects with storage more than 60 Mm3 are designed using flood frequency method with 1000-year return period (Subramanya, 2005).

Climate change is currently taking place due to elevated concentrations of ‘greenhouse gases’ in the atmosphere. According to the Intergovernmental Panel on Climate Change (IPCC), mean global surface temperature increased 0.6 ± 0.2 °C in the 20th century. Average sea level rose 0.1–0.2 m globally, and precipitation increased by 0.5–1% per decade, with an increase in the frequency of heavy precipitation events. Increases in annual and seasonal precipitation may increase the magnitude and frequency of flooding (Houghton et al., 2001). General Circulation Models (GCM), based on mathematical representations of atmosphere, ocean, ice cap and land surface processes, are considered to be the only credible tools currently available for simulating the response of the global climate system to increasing greenhouse gas concentrations. Accordingly, air temperature projected to increase by 1.4–5.8 °C and precipitation by 3–15% globally in the 21st century (Houghton et al., 2001, Swansburg et al., 2004).

Investigation of the hydrological impacts of climate change at the regional scale requires the use of a downscaling technique, which is a process of transferring the coarse spatial resolution of General Circulation Models (GCM) information to a form suitable for direct use in many types of climate impact models (Hashmi et al., 2011). Methods of downscaling were developed to obtain local-scale surface weather from regional-scale atmospheric variables that provided by GCMs. Two main forms of downscaling technique exist. One form is dynamical downscaling, where output from the GCM used to drive a regional, numerical model in higher spatial resolution, which therefore is able to simulate local conditions in detail. The other form is statistical downscaling, where a statistical relationship established from observations between large-scale variables, like atmospheric surface pressure, and a local variable, like the wind speed at a particular site. Statistical downscaling model facilitates the rapid development of multiple, low-cost, single-site scenarios of daily surface weather variables under current and future regional climate forcing (Wilby et al., 2002). Significant progress has already been made in the development of statistical downscaling techniques (Huth, 1999, Wilby et al., 2002, Hashmi et al., 2011).

Wilby and Wigley (1997) divided downscaling into four categories: regression methods, weather pattern-based approaches, stochastic weather generators, which are all statistical downscaling methods, and limited-area modeling. Among these approaches, regression methods are preferred because of its ease of implementation and low computation requirements. Regression based downscaling is the use of transfer function. The transfer function method relies on direct quantitative relationship between the local scale climate variable (predictand) and the variables containing the large-scale climate information (predictors) through some form of regression.

Scientific literature of the past decade contains a large number of studies regarding the development of downscaling methods and the use of hydrologic models to assess the potential effects of climate change on a variety of water resource issues. Hydrologic models provide a framework in which to conceptualize and investigate the relationship between climate and water resources (Xu, 1999). Precipitation, which is a key component of the hydrological cycle and one of the important parameters have taken the focus of noticeable scientists concerned with downscaling and climate change effects. The use of the downscaling in forecasting the precipitation concentrated in both the drought and flood cases. Precipitation downscaling improves the coarse resolution and poor representation of precipitation in global climate models, and helps end users to assess the likely hydrological impacts of climate change (Maraun et al., 2010).

A number of studies were conducted by using statistical downscaling and different GCM scenarios to predict the runoff based on precipitation and rainfall–runoff models (Schmidli et al., 2007, Chen et al., 2012, Yonggang et al., 2013). To the extent of the author’s knowledge, there are no studies conducted up to date using the observed flood discharge data directly for downscaling without the intermediate stage of using the rainfall–runoff models. Up-to-now, conventional statistical downscaling approaches linking GCMs to meteorologic and hydrologic models resolved at finer scales have been developed and implemented successfully, however, this is the state-of-the art approach to build a bridge between the hydroclimatic modeling at the river-basin fine-resolution scale and the coarse-resolution GCMs.

In this study, efforts is done to evaluate the possibility of using the discharge as predictand variable in the statistical downscaling, and as a second step efforts focused toward prediction of the peak monthly discharge which can be the base for calculation of design flood value on the basis of flood frequency analysis for the predicted discharge time series.

Section snippets

Study area and data

The first set of the data, which is the daily inflow data, were collected from Darbandikhan Dam. Darbandikhan Dam is a multi-purpose embankment dam on the Diyala River (locally called Sirwan River), 65 km southeast of Sulaymaniy Governorate, and 285 km northeast from Baghdad, Iraq (Coordinates 35°06′46″N 45°42′23″E). It was constructed between 1956 and 1961. The purpose of the dam is irrigation, flood control, hydroelectric power production, and recreation. The Diyala River is a tributary of the

Data pre-processing

The inflow data combined with the 26 downscaling inputs and the range used for NCEP set and CGCM3 set are the same as the available inflow data set (i.e. 36 years data starting from 11th March 1962 up to 10th March 1998). After combining all the data of each set with the inflow separately, the data is divided into training subset and validation (testing) subset. The data set was divided by taking 20% of the data randomly as validation subset and the rest as training subset. Preliminary trials of

Results and discussion

All the fourteen models ran under the same circumstances of run time, parameter settings, and computer hardware specification to have reasonable comparative basis. The coefficient of determination R2 and the Mean Squared Error (MSE) are considered in this study for model results comparison. Table 8 shows the value of R2 and MSE for each of the fourteen models. In general, the results shows that NCEP predictors have better correlation with the inflow data than the CGCM3 predictors do. For the

Conclusions

In this study, a new methodology for prediction of the design flood discharge is introduced. The historical flood discharge values of Darbandikhan dam built on Diyala River are predicted by statistical downscaling the CGCM variables based on linear genetic programming.

26 large-scale CGCM variables are used as input in the proposed downscaling methodology based on LGP regression technique to predict the measured local flood discharge. LGP has the ability of evaluating the auto-correlation among

Acknowledgements

We would like to thank to Scientific Research Projects Unit of Gaziantep University for supporting this study as a project (MF.12.11) by research funding.

References (25)

  • Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Xiaosu, D., 2001. Climate Change 2001: The...
  • R. Huth

    Statistical downscaling in central Europe: evaluation of methods and potential predictors

    Clim. Res.

    (1999)
  • Cited by (36)

    • Framework for incorporating climate projections in the integrated planning and management of urban infrastructure

      2022, Urban Climate
      Citation Excerpt :

      GCMs use emission scenarios, population projections, technology, and energy supply as inputs (Hayhoe et al., 2015). However, the GCMs are coarse resolution and can hardly be applied directly to studies on local and regional scales, requiring the use of dynamic and/or statistical downscaling approaches (Tofiq and Guven, 2014). In dynamic downscaling, regional climate models (RCM) are nested within the GCMs to downscale climatic projections, making complex design and computationally expensive.

    • From global circulation to local flood loss: Coupling models across the scales

      2018, Science of the Total Environment
      Citation Excerpt :

      Several studies reviewed downscaling methods for hydrological applications (e.g. Fowler et al., 2007; Kundzewicz and Stakhiv, 2010; Teng et al., 2012; Wilby, 2010). More recent studies focused on specific methodological problems in this procedure, namely on scale effects (Piniewski et al., 2013), rainfall statistics (Langousis et al., 2016), and hydro-meteorological extremes (Madsen et al., 2014; Tofiq and Guven, 2014; Sunyer et al., 2015). The results of these studies have confirmed the applicability of linking GCMs to hydrological models via RCMs.

    • Development of a method to identify change in the pattern of extreme streamflow events in future climate: Application on the Bhadra reservoir inflow in India

      2017, Journal of Hydrology: Regional Studies
      Citation Excerpt :

      Ashofteh et al. (2013a) has investigated and confirmed the climate change impact on monthly inflow volume of the reservoir in an East Azerbaijan river basin. Tofiq and Guven (2014) attempted to predict the peak monthly discharge from statistical downscaling approach. Mishra and Singh (2010) and Mishra et al. (2011) has investigated the changes in extreme precipitation in Texas.

    View all citing articles on Scopus
    View full text