A method and system for medium and long-term rainfall grading coupled forecasting based on statistical dynamics

Abstract

The invention discloses a method and system for medium- and long-term rainfall hierarchical coupling forecasting based on statistical dynamics, belonging to the field of rainfall forecasting, including: model training stage: collecting historical hydrometeorological data in target areas and global scales, and establishing statistical models and The medium and long-term rainfall forecast model library of dynamic models, for each of the rainfall forecast models, use the model to forecast the rainfall in different historical time periods, and divide the forecast results into k rainfall levels; based on the neural network The coupled forecasting model corresponding to each rainfall level is trained in combination with the measured rainfall in the same period; the rainfall forecasting stage: each rainfall forecasting model is used to forecast the rainfall in the target time period, determine the corresponding rainfall level, and select the corresponding The coupled forecasting model couples each forecasting result to obtain the final forecasting result. The invention can improve the accuracy of medium and long-term rainfall forecast and prolong the forecast period of rainfall forecast.

Description

technical field [0001] The invention belongs to the field of rainfall forecasting, and more specifically relates to a method and system for mid- and long-term rainfall hierarchical coupling forecasting based on statistical dynamics. Background technique [0002] Medium- and long-term hydrological forecast refers to the scientific prediction of hydrological elements in the future for a long period of time based on the previous meteorological and hydrological elements, based on the method of genetic analysis and mathematical statistics. Medium and long-term rainfall forecast can provide rainfall data for medium and long-term hydrological forecast, and is the basis of medium- and long-term hydrological forecast. The rainfall process in a watershed is a nonlinear time-varying process that strongly depends on dynamic characteristics such as atmospheric circulation, surface evaporation, and changes in global climate factors. It involves many variables and is subject to natural con...

AI Technical Summary

Problems solved by technology

[0003] The existing medium- and long-term rainfall forecasting methods mainly use specific models for rainfall forecasting. These methods can achieve medium- and long-term rainfall forecasting to a certain extent, but the accuracy of rainfall forecasting is usually limited by the quality of the input data, the structure of the selected forecast model and the optimization of the model. The parameters of the model are specifically reflected in the following aspects: ①Input aspect: atmospheric circulation, sea surface temperature, sea ice, snow cover and other meteorological factors are used as the main input information of the rainfall model, and the forecast period of rainfall forecast is limited by the timeliness of observed rainfall;② In terms of models: For the research area, a rainfall forecast model is usually used to simulate the rainfall process, ignoring the watershed adaptability of different rainfall forecast models, the forecast results are prone to large errors, leading to decision-making risks; ③ Parameter calibration: Optimal model The objective function of the parameters is single, and it is difficult to fully reflect the differences in the peak value, rainfall and evolution process of different rainfall processes in the watershed; ④ In terms of forecast accuracy: the guiding significance of the model forecast error to the forecast model is ignored, and the complementary characteristics of the model are not fully utilized. Correction of Forecast Results

[0004] In addition, the rainfall forecast of large watersheds has the characteristics of complex terrain and large spatial differences in the spatial dimension, and the characteristics of strong uncertainty in the time dimension, and is restricted by one or more of the above problems. The current rainfall forecast is difficult to meet Requirements for rainfall forecast accuracy and forecast period in watershed hydrological forecast, disaster prevention and mitigation, and water resources scientific management

Images