High-temporal-spatial-resolution rainstorm flood forecasting method and system for medium and small watershed

Abstract

The invention relates to rainfall and runoff forecasting techniques. The invention discloses a medium and small watershed high temporal-spatial resolution rainstorm flood forecasting method, which is characterized in that a WRF model is adopted as a rainfall forecasting mode, and a rainfall error of WRF forecasting is corrected through a deviation correction method; adopting a distributed hydrological physical model as a watershed hydrological model; based on the WRF model and the distributed hydrological physical model, establishing a rainstorm flood coupling forecasting model; and forecasting the basin flood by using the rainstorm flood coupling forecasting model. The technical bottleneck in rainstorm and flood forecasting in small and medium-sized watershed is broken through, the intelligence degree, accuracy, forecasting period and forecasting efficiency of rainstorm and flood coupling forecasting are improved, and the national major demand of flood control and disaster reduction and the wide application demand of the industry are met. An advanced and practical forecasting model platform is provided for production units such as meteorological bureau (meteorological observatory) and flood control office, and the method has important practical value, social significance and livelihood significance.

Description

Technical Field

[0001] The present application relates to the field of hydrological and meteorological technology, in particular to precipitation and runoff forecasting technology, and more specifically to a method and system for forecasting rainstorms and floods with high temporal and spatial resolution in small and medium-sized watersheds. Background Art

[0002] Precipitation data is an important model input for basin runoff forecasting, and its quality directly affects the accuracy of runoff forecast results. Early runoff forecasting used precipitation observed at rain gauges as input data for basin hydrological models. In theory, this observed precipitation data has the highest accuracy. However, precipitation observed at rain gauges is based on stations and cannot directly provide surface precipitation data in grid form. Due to the limitations of the density and spatial distribution of rain gauges, the grid-form precipitation data obtained after interpolation is uncertain, and it is difficult to accurately reproduce the true spatial distribution of precipitation, especially in small and medium-scale basins with insufficient data. Coupling numerical weather forecast models with basin hydrological models for runoff forecasting has become the focus of current hydrological and meteorological forecasting research. It is also one of the effective methods to solve the problems of insufficient data in small and medium-scale basins and the need for high temporal and spatial resolution of precipitation data. In constructing a meteorological-hydrological coupled runoff forecasting system based on the mesoscale numerical weather forecast model WRF and the high-resolution distributed hydrological forecast model BTOPMC, the advantages of the meteorological-hydrological coupled model, such as high temporal and spatial resolution and not being restricted by sparse observation data, are fully utilized.

[0003] With the development and progress of spatial information technology such as remote sensing, computer science and other related disciplines, distributed hydrological models, distributed hydrological physical models BTOPMC (Block-Wise use of TOPMEDEL with Muskingum-Cunge), and especially the coupled rainstorm-flood forecast (compared with hydrological models, it has unique potential advantages such as the ability to provide real-time forecast of rainstorm-flood linkage, which is expected to better improve the accuracy of flood forecasting and provide a longer flood forecast period), are one of the hot spots, difficulties and development trends in the field of basin-scale flood forecasting research.

[0004] The one-way coupling of rainstorm and flood only considers the impact of meteorological factors on the hydrological process, and uses the output field of the meteorological model as the driver of the hydrological model to realize the forecast of future runoff process. One-way coupling has the characteristics of high flexibility and convenient model debugging. Because it does not consider the feedback effect of the land surface model on the atmosphere, it is mainly used for medium- and short-term runoff forecasts. Since the rainstorm-flood process often lasts for a short time (generally a few hours to tens of hours), the feedback effect of the surface on the atmosphere such as evapotranspiration and radiation is relatively weak compared to the driving effect of the atmosphere on surface hydrological phenomena. Therefore, one-way coupling is generally used in flood forecasting.

[0005] In recent years, WRF (Weather Research and Forecasting Model) has become the fastest-growing mesoscale atmospheric model, and there are more and more coupled applications of the WRF model. Gao Bing et al. (2012) used the WRF model and the distributed hydrological model GBHM (Geomorphology-Based Hydrological Model), selected the Yangtze River Basin as the study area, and conducted a forecast experiment on the Three Gorges Reservoir flood. The predicted runoff process had a certain accuracy, and the land-air coupled runoff forecasting method had an obvious effect on extending the runoff forecast period. Tang et al. (2014) verified the effectiveness of the one-way coupling of the WRF model and the VIC model (Variable Infiltration Capacity) in evaporation, soil moisture, runoff and base flow simulation forecasting. Monteiro et al. (2013) input the 5km*5km hourly precipitation data output by the WRF model into the distributed hydrological model to forecast the runoff for the next 7 days, and based on this, optimized the power generation plan of small hydropower stations in mountainous areas, and achieved satisfactory results. Shih et al. (2014) combined the WRF model with WASH123D (WAterSHed systems of 1-D stream river network, 2-D overland regime, and The study found that the calibrated hydrological model could not adapt well to meteorological data and still caused large forecast errors. The reasons are: first, the accuracy of precipitation forecast is low, and second, the inconsistency between verification data and forecast driving data.

[0006] In general, due to the limitation of precipitation forecast accuracy, the accuracy of flood forecasting in small and medium-sized river basins using existing technologies is not high.

[0007] For flood forecasting in small and medium-sized river basins, it is reasonable to focus on the research and development of one-way coupled rainstorm-flood coupled forecasting theoretical methods. Due to the complexity of flood forecasting in small and medium-sized river basins and the historical limitations of related disciplines, such rainstorm-flood coupled forecasting models are still relatively rare. These basic key scientific issues urgently need to be studied and solved in depth so as to better serve the national needs and industry application needs for flood control and disaster reduction. Summary of the invention

[0008] The main purpose of this application is to provide a method and system for forecasting rainstorm and floods in small and medium-sized river basins with high temporal and spatial resolution, to correct the precipitation error of WRF forecast by means of deviation correction method, to provide high-precision precipitation data with a certain forecast period for flood forecasting, thereby achieving high-precision forecasting of floods, and solving the problem of low accuracy of flood forecasting in small and medium-sized river basins in the prior art.

[0009] In order to achieve the above-mentioned purpose, according to one aspect of the specific implementation method of the present application, a method for forecasting rainstorm and floods with high temporal and spatial resolution in small and medium-sized watersheds is provided, which is characterized in that the WRF model is used as the precipitation forecast mode, and the precipitation error of the WRF forecast is corrected by a deviation correction method; a distributed hydrological and physical model is used as the watershed hydrological model; a rainstorm and flood coupling forecasting model is established based on the WRF model and the distributed hydrological and physical model; and the watershed floods are forecasted by the rainstorm and flood coupling forecasting model.

[0010] In some embodiments, the precipitation error in the WRF forecast is corrected by a bias correction method, including a mean-based error correction method and a probability distribution-based error correction method.

[0011] In some embodiments, the probability distribution-based error correction method applies Delta quantile mapping to correct precipitation forecasts.

[0012] In some embodiments, during the error correction process, a transfer function between the observation and model data in a reference time period is constructed based on the cumulative probability distribution function of the two, and the transfer function is used to correct the cumulative probability distribution function of the simulation results in other time periods, thereby reducing the model simulation error.

[0013] In some embodiments, when correcting precipitation data, precipitation less than 0.1 mm / day in observations and model forecasts is assigned a value of 0, and the daily precipitation in the model results is sorted from large to small. After reaching the observed number of precipitation days, the remaining precipitation is assigned a value of 0, and finally a new daily precipitation sequence is obtained.

[0014] According to another aspect of the specific implementation of the present application, a high spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds is provided, including a WRF model, a distributed hydrological physics model, and a deviation correction module;

[0015] The WRF model is used as a basin precipitation forecast model;

[0016] The distributed hydrological physical model serves as a watershed hydrological model;

[0017] The distributed hydrological physical model includes a terrain module, a runoff generation module, a confluence module and an evapotranspiration module.

[0018] In some embodiments, the above-mentioned deviation correction module includes a mean-based error correction module and a probability distribution-based error correction module.

[0019] In some embodiments, the probability distribution-based error correction module applies Delta quantile mapping to correct precipitation forecasts.

[0020] In some embodiments, during the error correction process, a transfer function between the observation and model data in a reference time period is constructed based on the cumulative probability distribution function of the two; then, the transfer function is used to correct the cumulative probability distribution function of the simulation results in other time periods, thereby reducing the model simulation error.

[0021] In some embodiments, when correcting precipitation data, precipitation less than 0.1 mm / day in observations and model forecasts is assigned a value of 0, and the daily precipitation in the model results is sorted from large to small. After reaching the observed number of precipitation days, the remaining precipitation is assigned a value of 0, and finally a new daily precipitation sequence is obtained.

[0022] According to the technical solution of the present application and the further improved technical solution in certain exemplary embodiments thereof, the present application has the following beneficial effects:

[0023] It has broken through the technical bottleneck in rainstorm and flood forecasting in small and medium-sized river basins, improved the intelligence, accuracy, forecast period and forecast efficiency of rainstorm and flood coupling forecasting, met the major national needs for flood control and disaster reduction and the industry's wide application needs, and provided an advanced and practical forecasting model platform for production units such as the Meteorological Bureau (Weather Observatory) and the Flood Control Office, which has important practical value, social significance and livelihood significance.

[0024] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. Additional aspects and advantages of the present application will be partially given in the following description, partially become apparent from the following description, or be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings constituting a part of this application are used to provide a further understanding of this application. The specific implementation modes, illustrative embodiments and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0026] Figure 1 It is a structural schematic diagram of a forecast system according to a specific implementation method of the present application;

[0027] Figure 2 The figure is a flow chart of BTOPMC model calculation according to a specific implementation method of the present application. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the specific implementation modes, exemplary embodiments and features thereof in this application can be combined with each other. The present application will now be described in detail with reference to the accompanying drawings and in combination with the following content.

[0029] In order to enable those skilled in the art to better understand the solution of this application, the following will be combined with the drawings in the specific implementation methods and exemplary embodiments of this application to clearly and completely describe the technical solutions in the specific implementation methods and exemplary embodiments of this application. Obviously, the described exemplary embodiments are only part of the embodiments of this application, not all of them. Based on the specific implementation methods and exemplary embodiments in this application, all other implementation methods and embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application.

[0030] Example

[0031] Mesoscale Numerical Weather Forecast Model WRF

[0032] The mesoscale numerical weather forecast model used in this application is the WRF model, version 4.3, which is the current mainstream mesoscale weather forecast model. WRF is a new generation of mesoscale weather research and forecasting system jointly developed by the National Center for Atmospheric Research, the National Oceanic and Atmospheric Administration and other scientific research institutes. WRF is a fully compressible non-hydrostatic model, a model system that integrates numerical weather forecasting, atmospheric simulation and data assimilation. WRF is mainly composed of four parts: a preprocessing system (used to interpolate data and initialize model standards, define model areas, and select map projection methods), an assimilation system, a power kernel, and a post-processing part. The framework of the model (power kernel) is divided into two modules: Advanced Research WRF (ARW) and Nonhydrostatic Mesoscale Model (NMM), the former is mainly used for scientific research, and the latter is used for business forecasting.

[0033] The dynamic kernel used in the WRF model of this application is ARW. WRF is applicable to a range of meters to thousands of kilometers, and is mainly used in idealized experiments, sea-air coupling, meteorological case studies, regional climate research, global-scale simulation, and hydrology and water resources research. The initial field data processing process of this WRF can be summarized as follows: (1) Providing variables for initial field preprocessing, mainly including two-dimensional variables and three-dimensional variables, where two-dimensional variables also include static variables and dynamic variables that change over time; three-dimensional variables include: potential temperature, mixing ratio and momentum horizontal component; two-dimensional static variables include terrain, land use type (different classification standards can be selected), soil, vegetation cover, albedo, etc.; two-dimensional dynamic variables include soil moisture, sea ice, snow depth, etc.; (2) Determine the scope of the study area, resolution, number of grid points, and map projection method; (3) Interpolate the static field according to the user's set conditions; (4) Convert the format of the driving data (generally reanalysis data) to provide the initial field and boundary field for WRF operation; (5) Horizontal and vertical interpolation of the driving data, and finally complete the data initialization before the model runs.

[0034] The WRF model provides different physical parameterization schemes for different physical processes, including longwave radiation scheme, shortwave radiation scheme, cloud microphysical process scheme, boundary layer scheme, near-surface layer scheme, cumulus convection parameterization scheme and land surface process scheme.

[0035] Correction Method of Precipitation Numerical Forecast System

[0036] Error correction methods can be roughly divided into two categories: based on the mean and based on probability distribution. The Delta method is a more common mean-based correction method. Although it is simple and easy, it is generally only applicable to the correction study of the average state. It is difficult to apply to the processing of daily-scale data, and it cannot correct errors in probability distribution. The correction method based on probability distribution (also known as the quantile mapping method) can overcome this defect and is widely used. This application uses Delta quantile mapping (QDM, Quantile Delta Mapping) to correct precipitation forecast products. The core idea is to construct a transfer function between the observations and model data in the reference period during the error correction process. Then use this transfer function to correct the cumulative probability distribution function of the simulation results in other time periods, thereby reducing the model simulation error, and can be further applied to statistical downscaling.

[0037] When correcting the precipitation data, the precipitation less than 0.1 mm / day in the observation and model forecast is assigned a value of 0, and the daily precipitation in the model results is sorted from large to small. After reaching the number of days with observed precipitation, the remaining precipitation is assigned a value of 0, and finally a new daily precipitation sequence is obtained. The whole year is divided into four seasons: spring (March to May), summer (June to August), autumn (September to November), and winter (December to February). In the correction process, the precipitation observation data and model output data for 9 years in the simulation period from 2001 to 2009 are used to establish the transfer function, and the precipitation season from 2010 to 2015 is corrected. The precipitation less than 0.1 mm / day after correction is assigned a value of 0, and the corrected 4-year daily precipitation sequence is obtained and its effect is tested. Therefore, this chapter calls 2010 to 2015 the verification period, and all the evaluation and analysis in the article are also carried out for this period.

[0038] QDM (Quantile Delta Mapping) retains the changes in quantiles. The method described by Cannon et al. is similar to the equidistant and balanced forms of QM (Quantile Mapping). Detrending processing must be performed in each quantile first, and then the transfer function constructed during the modeling period is used to correct the data through the QM method, and the relative or absolute changes are superimposed on the corrected results.

[0039] For precipitation x, first calculate the non-excess probability associated with time t:

[0040]

[0041] In the formula, x m,p (t) refers to the model simulation value corresponding to the verification period time t, denoted by the subscript p, and the range of ε(t) is 0 to 1. is the pattern verification period x m,p The corresponding CDF (Cumulative Distribution Function).

[0042] For precipitation, the quantile choice between the modeling period and the validation period is expressed as a relative change, which can be defined as:

[0043]

[0044] Where F_(m,c)^(-1) is the inverse CDF obtained from the model simulation results during the modeling period. The simulated ε quantile can also be corrected in time period t during the modeling period based on the observed values, as follows:

[0045]

[0046] In the formula, is the value x from the observations during the modeling period o,c The inverse CDF calculated from .

[0047] Finally, in the future period t, the revised precipitation result x bc Equal to the error correction value Multiply by the relative change in precipitation Δ(t), as follows:

[0048]

[0049] Distributed Basin Hydrological Model BTOPMC

[0050] The BTOPMC model is a distributed watershed hydrological model based on physical mechanisms. It was originally developed by Ao Tianqi for distributed hydrological simulation of large watersheds during his study abroad at the University of Yamanashi, Japan. Currently, the BTOPMC model consists of a terrain module, a runoff module, a confluence module, and an evapotranspiration module.

[0051] In developing and studying the BTOPMC model, the researchers have achieved some results in river network technology, natural sub-basin division, spatial distribution of runoff model parameters in grid units, runoff accuracy, etc. The main contents involved are: (1) A new automatic analysis technology was developed to calculate the elevation micro-increment according to the spatial position of the depression in the Digital Elevation Model (DEM), thereby eliminating the depression; (2) The problem of using the Pfafstertter automatic watershed segmentation method for distributed hydrological models was studied and improvement measures were proposed; (3) A method for determining the degree of stable segmentation when using the Pfafstertter method to segment the watershed was proposed; (4) An approximate treatment method for solving the negative outflow problem in the Muskingum-Cunge runoff method and a real-time adjustment method for the time and space step were developed; (5) A new mode was proposed for using TOPMODEL for the runoff model of a large watershed, in which the model parameters are given by the watershed physical characteristics such as soil and vegetation in each grid; (6) The necessity of using the Pareto analysis method to reduce the uncertainty in the model parameter calibration process and the establishment of the parameter transfer function was proposed.

[0052] Construction of numerical prediction system for river basin precipitation

[0053] The precipitation numerical prediction system for this application is built based on WRFv4.3. The system area is set to a triple nested grid of 27, 9, and 3 km, with a time step of 150, 30, and 10 s, 50 layers in the vertical direction, and the top of the model layer is 50 hPa. The main physical process parameters in the system are set as: Thompson cloud microphysics scheme, Kain-Fritsch convection parameterization scheme, MYJ boundary layer scheme, Noah land surface scheme, RRTM longwave radiation and Dudhia shortwave radiation scheme. The model uses the GFS (Global Forecast System) data issued by the National Center for Environmental Forecast (NCEP) of the United States as the initial field, and the forecast period is 36 hours.

[0054] The evaluation indicators of precipitation forecast accuracy are divided into two categories, statistical indicators and detection indicators. The formulas and optimal values ​​of each indicator are shown in Table 1. Among them, the statistical indicators include correlation coefficient CC, mean absolute deviation MAE, mean relative deviation MRE, relative standard deviation RSD, root mean square error RMSE and standardized root mean square error NRMSE, which are used to evaluate the correlation and deviation degree between the precipitation forecast product and the observed information in terms of numerical value. CC reflects the strength of the linear relationship between the precipitation forecast product and the observed precipitation information. The closer the absolute value of CC is to 1, the more consistent the precipitation forecast product is with the ground observed precipitation information, and the higher the reference value. MAE and MRE describe the degree of deviation between the precipitation forecast product and the observed value, and MRE includes positive deviation (overestimation) and negative deviation (underestimation). The closer the deviation is to 0, the closer the precipitation product is to the true value. RMSE also measures the degree of deviation between the forecast value and the observed value. The smaller the RMSE, the smaller the observation error, and vice versa.

[0055] Table 1 Precipitation data accuracy evaluation indicators, formulas and optimal values

[0056]

[0057] In Table 1, n represents the number of data pairs used in the accuracy evaluation. Si and Oi represent the WRF forecast precipitation value and the observed value at the i-th grid point, respectively. and Represent the average WRF forecast precipitation and the average precipitation observation of all grid points, respectively. TP represents the number of precipitation events that are successfully captured by both WRF forecast precipitation and observation at a specific threshold. FN represents the number of precipitation events that are successfully captured by observation but not by WRF forecast precipitation at a specific threshold. FP represents the number of precipitation events that are captured by WRF forecast precipitation but not by observation at a specific threshold.

[0058] The detection indicators include the critical success index CSI, BIAS score, hit rate POD, false alarm rate FAR, accuracy rate Precision and F score Fscore, which are used to evaluate the precipitation forecast product's ability to capture precipitation events. POD represents the proportion of correctly detected events to the total number of detected events among the precipitation events detected by the precipitation product, reflecting the degree of underreporting of precipitation events by the precipitation product. FAR reflects the proportion of falsely detected events to the total number of detected events among the precipitation events detected by the precipitation product, reflecting the degree of false reporting of precipitation events by the precipitation product. CSI represents the proportion of correctly detected precipitation events to the total number of events by the precipitation product, which can comprehensively reflect the characteristics of the precipitation product. Precision is the probability that an event is correctly detected as rainfall. Fscore combines the characteristics of accurate detection, false detection and missed detection. The calculation of detection indicators depends on the threshold setting for precipitation events. This application sets five categories of daily precipitation events and analyzes the precipitation detection capabilities under different precipitation intensities (mm / day): 0.1-1 (trace precipitation); 1-5 (small precipitation); 5-10 (medium precipitation); 10-20 (large precipitation); >20 (heavy precipitation).

[0059] Correction and effect evaluation of precipitation numerical forecast system

[0060] Since ground observation stations are sparsely distributed on the Qinghai-Tibet Plateau, gridded precipitation data are used to evaluate the precipitation products predicted by the numerical prediction system. The reference observation values ​​of this application use the China Regional Surface Meteorological Element Dataset (CMFD) developed by the Institute of Tibetan Plateau Research, Chinese Academy of Sciences. Due to its high accuracy, it has been widely used in precipitation research on the Qinghai-Tibet Plateau. The CMFD dataset is based on the existing Princeton reanalysis data, GLDAS meteorological driving data, GEWEX-SRB radiation data and TRMM precipitation data in the world. It is produced by integrating the conventional meteorological observation data of the China Meteorological Administration. The horizontal resolution is 0.1°×0.1°. The overall accuracy of the data is high and has been widely used in hydrological and climate simulation research. Wen Tingting et al.

[19] used the measured precipitation data of 131 meteorological stations on the Qinghai-Tibet Plateau from 1979 to 2017 to evaluate the applicability of three reanalysis precipitation data, CRU, ERA5 and CMFD, on the Qinghai-Tibet Plateau at the annual and seasonal scales. The results show that the precipitation products of CMFD are better than ERA5 and CRU on the Qinghai-Tibet Plateau. Huang Qi et al.

[20] used CMFD and ground-based precipitation as input to drive the Xinanjiang model. The results showed that the CMFD precipitation simulation results were better than the station precipitation simulation results and could better simulate runoff.

[0061] The revised method shows satisfactory improvement effects on the precipitation probability, daily precipitation, monthly precipitation, annual and seasonal precipitation, and extreme precipitation forecasted by WRF.

[0062] Construction of meteorological and hydrological coupled forecasting model

[0063] The climate model prepares the corresponding numerical forecast results according to the time period and spatial resolution required by the hydrological model, but there is a problem of mismatch between the climate model output and the hydrological model input format during the coupling process. The precipitation format output by the climate model is the Netcdf storage format, while the input format required by the hydrological model is text data ending with *.precpnt, *.precsit and *.prec, which respectively store the code and longitude and latitude of each precipitation grid, the spatial distribution of the grid and the corresponding time series precipitation. Therefore, it is necessary to add a conversion interface to control the coupling in BTOPMC, so that the precipitation output format of the WRF model can be directly used in the BTOPMC model for runoff forecasting after conversion, and realize automatic calibration and parameter adjustment of the hydrological model to obtain simulation results, so that the entire coupling model is automated, highly flexible and easy to debug. The basic framework of the overall coupling model is as follows: Figure 1 As shown in the figure, SCE-UA represents a hybrid complex evolutionary algorithm for parameter calibration; WPS is the abbreviation of WRF Preprocessing System. The coupled model mainly includes the WRF model, the distributed hydrological physics model (BTOPMC) and the deviation correction module. The output runoff forecast results (flood forecast results) have an average Nash efficiency of ≥85% for flood forecast accuracy under the condition of 1km-1hr resolution, and the forecast period can reach 72 hours (3 days).

[0064] Data required for the model

[0065] The data required for the model include meteorological data, land use data, leaf area index, soil data, topographic data and hydrological data. The watershed digital elevation model (DEM) data comes from the Shuttle Radar Topography Mission (SRTM), using SRTM3 topographic data with a resolution of 90m, which can be downloaded from the Geospatial Data Cloud website (http: / / www.gscloud.cn / ). The land use data comes from the MCD12Q1 product in MODIS, and the leaf area index comes from the MCD15A2H product, with a temporal resolution of 16d and a spatial resolution of 500m. The soil data comes from the Food and Agriculture Organization of the United Nations (FAO). Meteorological data include precipitation, temperature, evapotranspiration, etc. The temperature data comes from meteorological station observations, and the evapotranspiration data comes from the CRU monthly product. The precipitation data comes from the output of the WRF forecast model. To ensure the consistency of the input spatial resolution, all data are resampled to 1km by the nearest neighbor sampling method.

[0066] This application simulates the runoff of Lhasa and Niyang River basins. The runoff data are daily runoff data from 2010 to 2015, including Lhasa Station, Gengzhang Station, Baheqiao Station and Gongbujiangda Station. The runoff data from 2010 to 2013 are selected for model parameter calibration, and the data from 2014 to 2015 are used for model verification.

[0067] Model evaluation metrics

[0068] The Nash efficiency coefficient (NSE), correlation coefficient (CC) and root mean square error are used to evaluate the model results. The closer the NSE and CC are to 1, the better the simulation effect is, and the closer the RMSE is to 0, the higher the accuracy is. The calculation formulas for each indicator are as follows:

[0069]

[0070]

[0071]

[0072] Where n is the total number of samples, Qs is the simulated runoff, and Qo is the observed runoff.

[0073] The calculation flow chart of BTOPMC model is as follows Figure 2 shown.

Claims

1. A high temporal and spatial resolution rainstorm and flood forecasting method for small and medium-sized watersheds, characterized by: The WRF model is used as the precipitation forecast model, and the precipitation error of the WRF forecast is corrected by the deviation correction method. The distributed hydrological physics model is used as the basin hydrological model. Based on the WRF model and the distributed hydrological physics model, a rainstorm and flood coupling forecasting model is established; and the basin flood is forecasted by the rainstorm and flood coupling forecasting model.

2. The method for predicting rainstorm and flood with high temporal and spatial resolution in small and medium-sized watersheds according to claim 1 is characterized in that: The precipitation error of WRF forecast is corrected by bias correction methods, including error correction method based on mean and error correction method based on probability distribution.

3. The method for predicting rainstorm and flood with high spatiotemporal resolution in small and medium-sized watersheds according to claim 2 is characterized in that: The error correction method based on probability distribution applies Delta quantile mapping to correct precipitation forecasts.

4. The method for predicting rainstorm and flood with high spatiotemporal resolution in small and medium-sized watersheds according to claim 3 is characterized in that: In the error correction process, based on the cumulative probability distribution functions of observation and model data in the reference period, a transfer function between the two is constructed, and the cumulative probability distribution functions of simulation results in other periods are corrected using this transfer function, thereby reducing the model simulation error.

5. The method for predicting rainstorm and flood with high temporal and spatial resolution in small and medium-sized watersheds according to claim 3 is characterized in that: When correcting the precipitation data, the precipitation less than 0.1 mm / day in the observation and model forecast is assigned a value of 0, and the daily precipitation in the model results is sorted from large to small. After reaching the observed number of precipitation days, the remaining precipitation is assigned a value of 0, and finally a new daily precipitation sequence is obtained.

6. High spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds, characterized by: Includes WRF model, distributed hydrological physics model and deviation correction module; The WRF model is used as a basin precipitation forecast model; The distributed hydrological physical model serves as a watershed hydrological model; The distributed hydrological physical model includes a terrain module, a runoff generation module, a confluence module and an evapotranspiration module.

7. The high spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds according to claim 6 is characterized in that: The above-mentioned deviation correction module includes an error correction module based on a mean value and an error correction module based on a probability distribution.

8. The high spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds according to claim 7 is characterized in that: The error correction module based on probability distribution applies Delta quantile mapping to correct the precipitation forecast.

9. The high spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds according to claim 8 is characterized in that: In the error correction process, a transfer function between the observation and model data in the reference period is constructed based on the cumulative probability distribution function of the two; then, the transfer function is used to correct the cumulative probability distribution function of the simulation results in other periods, thereby reducing the model simulation error.

10. The high spatiotemporal resolution rainstorm and flood forecasting system for small and medium-sized watersheds according to claim 7 is characterized in that: When correcting the precipitation data, the precipitation less than 0.1 mm / day in the observation and model forecast is assigned a value of 0, and the daily precipitation in the model results is sorted from large to small. After reaching the observed number of precipitation days, the remaining precipitation is assigned a value of 0, and finally a new daily precipitation sequence is obtained.

Images