System and method for forecasting floods

Abstract

A method for forecasting flood, the method including: calibrating the hydrological model by using an objective function that is a sum of squared difference between the observed streamflow and the corresponding forecasted streamflow at each lead time to obtain the optimized hydrological model; using the optimized hydrological model to forecast floods; and evaluating forecasting performance of the optimized hydrological model. The method improves the forecasting accuracy and provides forecasting results at various lead times.

Claims

1. A method for forecasting flood, the method comprising: 1) obtaining hydrological variables comprising precipitation and evaporation for a basin, wherein the precipitation is obtained by measuring rainfall data with radars, rain gauges, and precipitation micro-physical characteristics sensors (PMCSs) and by combining the rainfall data from the radars, the rain gauges, and the PMCSs; 2) collecting historical flood information for the basin to estimate a mean concentration time of the basin and to determine a maximum lead time k; 3) obtaining a series of observed streamflow Q.sub.t.sup.obs, wherein the observed streamflow Q.sub.t.sup.obs represents the observed streamflow at a time t; t is an integer, and 1tN; and N is a total number of the observed streamflow; 4) establishing a hydrological model, and using the hydrological model for prediction by inputting the hydrological variables to the hydrological model to obtain a series of forecasted streamflow Q.sub.t,t-j.sup.pre; wherein the forecasted streamflow Q.sub.t,t-j.sup.pre represents the forecasted streamflow at the time t at a lead time j; j is an integer, and 1jk; and the precipitation within lead times are neglected for the prediction; 5) calibrating parameters of the hydrological model with an objective function as follows to obtain the optimized hydrological model: $\min F=\underset{t-1}{\overset{N}{.Math.}}\left({\left(Q_t^{\mathrm{obs}}-Q_{t,t-1}^{\mathrm{pre}}\right)}^2+{\left(Q_t^{\mathrm{obs}}-Q_{t,t-2}^{\mathrm{pre}}\right)}^2+\mathrm{.Math.}+{\left(Q_t^{\mathrm{obs}}-Q_{t,t-k}^{\mathrm{pre}}\right)}^2\right);$ 6) using the optimized hydrological model for prediction to obtain results for flood forecasts at multiple lead times, and sending the results for flood forecasts at the multiple lead times to manager computing devices for flood prevention by evaluating the forecasting performance of the optimized hydrological model using at least one criterion selected from: a Nash-Sutcliffe efficiency, a root mean square error, a water balance index, a qualified rate of peak flow, and a qualified rate of a peak time.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a flow chart of the method for forecasting flood at multiple lead times according to one embodiment of the invention;

(2) FIG. 2 is a schematic diagram of comparison between the observed results, the forecasting results of the invented method at 1-day lead time, and the forecasting results of the conventional method at 1-day lead time;

(3) FIG. 3 is a schematic diagram of comparison between the observed results, the forecasting results of the invented method at 2-day lead time, and the forecasting results of the conventional method at 2-day lead time; and

(4) FIG. 4 is a schematic diagram of comparison between the observed results, the forecasting results of the invented method at 3-day lead time, and the forecasting results of the conventional method at 3-day lead time.

DETAILED DESCRIPTION OF THE EMBODIMENTS

(5) For further illustrating the invention, examples detailing the method and system for flood forecasting are further set forth below. It should be noted that the following examples are intended to describe and not to limit the invention.

(6) As shown in FIG. 1, the method for forecasting flood at multiple lead times is as follows:

(7) obtaining hydrological variables comprising precipitation and evaporation for a basin, in which rainfall data is measured with radar, rain gauges, and the precipitation micro-physical characteristics sensor (PMCS), and the rainfall data from the three sources are combined to calculate the precipitation;

(8) collecting historical flood information to estimate the average time of concentration for the basin so as to determine the length of lead time (i.e., the maximum lead time k for flood forecasting);

(9) establishing a proposed hydrological model and using the proposed hydrological model for prediction to obtain the forecasted streamflow by inputting the hydrological variables (i.e. precipitation and evaporation) to the hydrological model;

(10) calibrating the hydrological model to obtain the optimized parameters for the model so as to result in the optimized hydrological model by using the following objective function, which is sum of the squared difference between the observed streamflow at each time and the forecasted streamflow at the corresponding time at every lead time within the maximum lead time k;

(11) $\begin{array}{cc}\min F=\underset{t-1}{\overset{N}{.Math.}}\left({\left(Q_t^{\mathrm{obs}}-Q_{t,t-1}^{\mathrm{pre}}\right)}^2+{\left(Q_t^{\mathrm{obs}}-Q_{t,t-2}^{\mathrm{pre}}\right)}^2+\mathrm{.Math.}+{\left(Q_t^{\mathrm{obs}}-Q_{t,t-k}^{\mathrm{pre}}\right)}^2\right);& \left(1\right)\end{array}$

(12) in the model calibration, .sup.obs is the observed streamflow at time t (t=1, 2, . . . , N), Nis the total number of the observed streamflow; Q.sub.t,t-j.sup.pre is the forecasted streamflow at time t at a lead time j, which is based on the inputs for forecasting at time j ahead of time t;

(13) using the optimized hydrological model to obtain forecasted results and sending the forecasted results to devices of managers for flood prevention; and

(14) validating the optimized hydrological model to evaluate the forecasting performance of the model by using a widely used criterion selected from: the Nash-Sutcliffe efficiency (NSE), Root Mean Square Error (RMSE), Water balance index (WBI), the qualified rate of peak flow (QRF) and the qualified rate of peak time (QRT).

(15) In the method, the parameters of the model are calibrated not only by using a single optimized algorithm but also by various combined algorithms. For example, the estimated parameters of Genetic algorithm can be treated as the initial values of the Rosen Brock methods, as well as the estimates of Rosen Brock methods can treated as the initial values of Simplex method.

(16) The formulas for the model validation are as follows:

(17) $\begin{array}{cc}\mathrm{NSE}=1-\frac{\underset{t=1}{\overset{N}{.Math.}}{\left(Q_t^{\mathrm{pre}}-Q_t^{\mathrm{obs}}\right)}^2}{\underset{t=1}{\overset{N}{.Math.}}{\left(Q_t^{\mathrm{obs}}-\stackrel{\_}{Q_{\mathrm{obs}}}\right)}^2};& \left(2\right)\\ \mathrm{RMSE}=\sqrt{\frac{\underset{t=1}{\overset{N}{.Math.}}{\left(Q_t^{\mathrm{pre}}-Q_t^{\mathrm{obs}}\right)}^2}{N}};& \left(3\right)\\ \mathrm{QRF}=\frac{\mathrm{NF}}{M};& \left(4\right)\\ \mathrm{QRT}=\frac{\mathrm{NT}}{M};& \left(5\right)\end{array}$
in which, .sup.pre is the predicted streamflow at time t; is the mean value of the observed streamflow; W.sub.pre and W.sub.obs are the total volume of the predicted and observed flow, respectively; NF is the number of the qualified flood events about peak flow; NT is the number of the qualified flood events about peak time; M is the total flood events.

Example

(18) The method of the invention is implemented with respect to Xunhe river basin which is located at the Shanxi province. The basin has a drainage area of 6448 km2, a length of the river of approximately 218 km, and the average annual flow of about 73 m.sup.3/s. The basin has a subtropical monsoon climate which is wet and moderate with an annual average temperature of 15-17 C. The rainfall across the basin in summer is abundant, which accounts for a large proportion (70%-80%) of the yearly rainfall.

(19) In the method, 28 rainfall stations and 3 hydrologic gauged stations are spreading over the basin for data collection. The areal precipitation is collected by applying the Thiessen polygon method to the gauge data of the rainfall stations, and the areal pan evaporation is computed by using the average value of the data from gauged stations. Discharge is measured at the outlet of the basin. Across the period 1991-2001, data of 1991 is used for warm-up. Data during 1992-1996 and 1997-2001 are served for calibration and validation, respectively.

(20) The invented method and the conventional method are both used for the flood forecasting at lead times of 1-3 days. The NSE, RMSE and WBI are used as the evaluation indicators for model validation.

(21) TABLE-US-00001 TABLE 1 Evaluation of flood forecasting performance of the conventional and invented methods Lead Calibration period Validation period time (day) NSE WBI RMSE NSE WBI RMSE Invented 1 0.90 1.14 61.24 0.79 1.17 62.36 method 2 0.62 0.90 119.09 0.57 0.94 90.16 3 0.33 0.70 157.82 0.24 0.75 120.42 Conventional 1 0.94 0.96 46.36 0.88 0.94 47.66 method 2 0.51 0.69 135.50 0.43 0.71 103.71 3 0.23 0.56 169.49 0.14 0.60 127.39

(22) The performances of conventional and invented methods are summarized in Table 1 as above. As shown in Table 1, the forecasting accuracy of both methods decreases with the increased lead time. In both the calibration period and the validation period, the NSE in the invented method is smaller than that in the conventional method for 1-day lead time, but the NSE in the invented method is larger than that in the conventional system for both 2-day and 3-day lead times. Similarly, the invented method has significant improvements in terms of WBI and RMSE as compared to the conventional method for the 2-day and 3-day lead times.

(23) FIGS. 2-4 demonstrate the comparison between the observed streamflow, the forecasting streamflow of the invented method, the forecasting streamflow of the conventional method from 1992/1/12 to 1994/3/22. FIGS. 2-4 show that the forecasting performance of the invented method is better than that of the conventional method for 2-day and 3-day lead times, especially in terms of the simulation of peak flows. In addition, FIGS. 2-4 show that compared with the conventional method, the performance of the invented method decreases slower as the lead time increases, especially in terms of the simulation of peak flows.

(24) Unless otherwise indicated, the numerical ranges involved in the invention include the end values. While particular embodiments of the invention have been shown and described, it will be obvious to those skilled in the art that changes and modifications may be made without departing from the invention in its broader aspects, and therefore, the aim in the appended claims is to cover all such changes and modifications as fall within the true spirit and scope of the invention.