Optimal Reservoir Operation Using Quantitative Precipitation Forecast During Typhoon Invasion
Typhoons in South East Asia bring rainfall expected for agriculture, but their excess might also cause floods. In order to reduce the damages effectively, a system is needed to support reservoir operators in flood control taking advantage of the available quantitative precipitation forecast. The spatial variability of the quantitative forecast is captured by a distributed hydrological model with an embedded reservoir operation. To get the optimal decisions a heuristic algorithm drives the hydrological model evaluating different release combination sets based on stochastic seeding. The error forecast is evaluated and introduced into the iteration as a weight to the release of each reservoir. The proposed system was applied to the upper Tone River in Japan with three multipurpose reservoirs. The efficiency was evident in reducing the flood peaks and volume downstream. Also, reservoirs' water levels at the end of the operation showed the potential for hydropower and agriculture. This approach shows the feasibility to be used as a real-time reference for flood management.
Bayesian MCMC Bandwidth Estimation on Kernel Density Estimation for Flood Frequency Analysis
Recent advances in computational capacity allow the use of more sophisticated approaches that require high computational power, such as the technique of importance sampling and Bayesian Markov Chain Monte Carlo (BMCMC) methods. In flood frequency analysis, the use of BMCMC allows to model the uncertainty associated to quantile estimates obtained through the posterior distributions of model parameters. BMCMC models have been used in association with various parametric distributions in the estimation of flood quantiles. However, they have never been applied with nonparametric distributions for the same objective. In this paper, the BMCMC is used for the selection of a bandwidth of the kernel density estimate (KDE) in order to carry out extreme value frequency analysis. KDE has not gained much acceptance in the field of frequency analysis because of its feature of easily dying off away from the occurrence point (low predictive ability). The use of Gamma kernels allow to solve this problem because of the caracteristics of thicker right tails and variable kernel smoothness. Even if the bandwidth is not changed, the gamma kernel permits alternating its variance according to the estimate point. Furthermore, BMCMC provides the uncertainty induced from the bandwidth selection. The predictive ability of the Gamma KDE is investigated with Monte Carlo simulation. Results show the usefulness of the gamma kernel density estimate in flood freuquency analysis.
Evaluation of Climate Change Hydrological Impacts and Associated Uncertainty Over the Mackenzie River Basin Using the CRCM
Fresh water discharge from the Mackenzie River Basin into the Arctic Ocean is an important component of the high-latitude hydroclimatic system. This northern river is characterized by a large drainage area of about 1.7 x 106 km2 and by relatively sparse surface observations. A great international effort during the Mackenzie GEWEX Study (MAGS) has been made to quantify and close water budget over the basin. Quantifying key components of the water budget and bringing together multiple information sources such as surface and atmospheric observations, reanalyses and different model simulations has been crucial for better understanding of the basin's hydrological behavior under the present climate. A comprehensive projection of future hydrological behavior of the basin also requires consideration of the entire water cycle and should be based on ensembles of climate change simulations. The Canadian Regional Climate Model (CRCM) simulates many of the complex processes in the hydrological cycle and can generate quantitative information of both atmospheric and terrestrial water budget components. In this study, an ensemble of eight CRCM simulations performed over the North-American domain (referred to as AMNO), at a 45-km resolution, under the A2 scenario is analyzed to obtain a comprehensive projection of the changes in both atmospheric and terrestrial water budget components over the Mackenzie River Basin. An evaluation of simulated water budget components with available observations under present climate conditions was also included in the analysis.
Comparison of Downscaled RCM and GCM Data for Hydrologic Impact Assessment
From observations of increases in global average air and ocean temperatures, widespread melting of snow and ice, and significant increases in net anthropogenic radiative forcing, it is clear that our global climate system is undergoing substantial warming (IPCC, 2008). A key area of concern for hydrologists and engineers alike is to determine how this warming will affect various hydrologic processes. To date, climate change impact studies have generally involved the downscaling of large-scale atmospheric predictors with the result then being input into a hydrological model to see how the flow regimes in a river/basin will change under various future climate change scenarios. Although many studies have been completed using large scale global climate model (GCM) simulations, few studies have shown the benefits of regional climate models (RCM). In this work, comparisons are made between the effectiveness of using CRCM4.2 vs. CGCM3.1 data as well as using statistically downscaled inputs in a climate change impact study. The study area is the Chute-du-Diable sub-basin located within the Saguenay-Lac-Saint-Jean Watershed in Quebec, Canada. Downscaled results are compared with observed meteorological data for the years 1961-1990 at the Chute- des-Passes (CDP) and Chute-du-Diable (CDD) weather stations; and flow is simulated in the Mistassibi River and the Chute-du-Diable reservoir. A statistical based regression technique (SDSM) and a dynamic artificial neural network model (Time lagged feed-forward neural network (TLFN)) are used for downscaling both the CRCM4.2 and CGCM3.1 data, and the HBV2005 hydrological modeling system is used for simulating flows in the watershed. For the base-line period (1961-1990), downscaling results revealed that downscaled CRCM4.2 precipitation and temperature series are closer to observed meteorological data at both CDD and CDP stations than downscaled CGCM3.1 series. The Wilcoxon Rank-Sum test and Levene test revealed that although TLFN and SDSM are capable of capturing the monthly means for precipitation and temperature accurately while SDSM is much better at capturing the degree of variability than TLFN. Statistical analysis also revealed that TLFN is best for downscaling temperature while SDSM is best for downscaling precipitation. With respect to the future (SRES A2) climate scenario both SDSM and TLFN revealed an 8-30% increase in precipitation and 1.8-5oC increase in mean temperature by 2050s with CGCM3.1 showing a larger increase than the CRCM4.2 model. Moreover, hydrologic simulations based on both statistically and dynamically downscaled precipitation and temperature inputs show increases in river flow and reservoir inflow throughout all seasons except for the summer where reduction of flow is observed. Annually, mean flow increases by about 16% to 45% in the 2050s with CRCM4.2 showing smaller increases than CGCM3.1. The study results also revealed that employing downscaled RCM data can yield better results than raw RCM and downscaled GCM data. The study results suggest that statistically downscaling RCM data could improve hydrologic impact assessment results at the catchment scale.
Assessment of the total predictive uncertainty of a real-time hydro-meteorological flood forecasting system using bivariate meta-gaussian density.
Medium range rainfall forecasts are increasingly used in operational flood forecasting applications as they provide an inviting option for extending prediction lead-times. Nonetheless, there is significant uncertainty associated with hydro-meteorological simulations. As a matter of fact, techniques for assessing hydro- meteorological model uncertainty have received a great deal of attention by researchers in recent years. In any flood forecasting system, the predictive uncertainty originates from several causes interacting between each other, namely input uncertainty, model structure uncertainty and parameter uncertainty. Furthermore, it appears to be difficult to isolate the errors that stem from the individual model components. In this framework, the study focuses on the analysis of the statistical properties of deterministic hydro- meteorological model error series, computed with respect to historic time series of observed discharge, in order to provide confidence intervals of discharge forecasts. Based on model error statistics, the proposed approach leads to the estimation of the uncertainty in an aggregated system (coupled atmospheric-hydrologic models), thereby rendering the assessment of uncertainty originating from the individual contributions unnecessary. Nevertheless, it is difficult to infer statistical properties from the prediction error since the residuals often appear to be non-stationary, in particular heteroscedastic, affected by serial correlation and with a non normal distribution. To solve this problem, the estimation of probability distributions of runoff simulation errors, conditioned by the value of flow, is performed using a meta-gaussian model. The latter is based on the application of a standard Normal Quantile Transform that makes the distribution of the model outputs and the model errors Gaussian in order to render straightforward the computation of confidence intervals. The approach is tested by the means of a case study that focuses on a real-time flood forecasting system that was set-up on the Alzette River in Luxembourg. The integrated flood forecasting system uses the rainfall and temperature forecasts of the American atmospheric GFS model (deterministic run) as forcing data in a conceptual hydrological model (deterministic run) to predict river discharge. Confidence intervals of discharge forecasts are computed for various prediction lead times and compared with the respective observations of river discharge.
Calibration of a Hydrologic Model Considering Input Uncertainty in Assessing Climate Change Impact on Streamflow
Studies on impact assessment and the corresponding uncertainties in hydrologic regime predictions is of paramount in developing water resources management plans under climate change scenarios,. The variability in hydrologic model parameters is one of the major sources of uncertainties associated with climate change impact on streamflow. Uncertainty in hydrologic model parameters may arise from the choice of model calibration technique, model calibration period, model structure and response variables. The recent studies show that consideration of uncertainties in input variables (precipitation, evapotranspiration etc.) during calibration of a hydrologic model has resulted in decrease in prediction uncertainty. The present study has examined the significance of input uncertainty in hydrologic model calibration for climate change impact studies. A physically distributed hydrologic model, Soil and Water Assessment Tool (SWAT), is calibrated considering uncertainties in (i) model parameters only, and (ii) both model parameters and precipitation input. The Markov chain Monte Carlo algorithm is used to estimate the posterior probability density function of hydrologic model parameters. The observed daily precipitation and streamflow data of the Canard River watershed of Essex region, Ontario, Canada are used as input and output variables, respectively, during calibration. The parameter sets of the 100 most skillful hydrologic model simulations obtained from each calibration technique are used for predicting streamflow by 2070s under climate change conditions. In each run, the climate predictions of the Canadian Regional Climate Model (CRCM) for SRES scenario A2 are used as input to the hydrologic model for streamflow prediction. The paper presents the results of uncertainty in seasonal and annual streamflow prediction. The outcome of the study is expected to contribute to the assessment of uncertainty in climate change impact studies and better management of available water resources.
CCSM Modeling for studying the impact of wetland drainage on hydro-climatology of the Midwest USA
Modeling experiments are performed to investigate the hydro-climatic impact of century long drainage in the United States Midwest region using National Center for Atmospheric Research's Community Climate System Model (CCSM Version 3.0). Coupled land surface and atmospheric components of CCSM are used at T85 spatial resolution (140 km) to create four control model experiments: two including wetlands and two excluding wetlands. The spatial distribution of wetland drained area, which accounts for 20% of the total area, is obtained from United States Agricultural Census reports. Four model control cases include: (i) 355 ppm CO2 for year 1990 excluding wetland (existing conditions), (ii) 355 ppm CO2 for 1990 including wetland, (iii) 289 ppm CO2 for year 1870 excluding wetland, and (iv) 289 ppm CO2 including wetland. Verification of CCSM results for first control case using North American Regional Reanalysis (NARR) data show that temperature is well predicted by CCSM, whereas there is large under estimation of summer and fall precipitation. Similarly, sensible heat flux is over estimated during summer and fall; whereas latent heat flux is underestimated during these seasons. Comparison of results from control cases one and two show pronounced effect of wetland drainage on sensible and latent heat fluxes. Reduction in wetland area for the 1990 control case show significant increase in sensible heat flux, and decrease in latent heat flux for most part of the year. Similarly, an increase in temperature and decrease in convective precipitation during summer months is observed. Results from 1990 control runs are also compared with 1870 control runs to discuss the effects of wetland drainage and green house gas emission on hydrologic budget (fluxes and storages) of the Midwest region.