Regional drought and flood distribution prediction method and system based on big data and multiple models

Abstract

The invention discloses a regional drought and flood distribution prediction method and system based on big data and multiple models. Historical rainfall and circulation sea temperature field data of a research area are collected, historical samples of typical drought and flood years are screened, SPI indexes of the typical drought and flood years are calculated, and time and space characteristics of the typical drought and flood years are obtained; constructing a typical drought and flood annual conceptual physical model and a regression clustering forecasting physical model forecasting map, and obtaining a climate background abnormal characteristic value and a drought and flood grade forecasting value of the research area; constructing a precipitation field optimization prediction model based on a singular value decomposition method; and constructing a drought and flood distribution prediction model of the research area based on the improved RegCM4 model to obtain drought and flood distribution of the research area. Compared with a traditional drought and flood prediction method, the method can provide a technical basis for drought and flood long-term rainfall prediction in the future, and improves the long-term rainfall prediction level.

Description

Technical Field

[0001] The present invention relates to a regional drought and flood distribution prediction method, in particular to a regional drought and flood distribution prediction method based on big data and multiple models. Background Art

[0002] Under the background of global warming, the changes in the temporal and spatial distribution of precipitation have led to increasingly frequent extreme weather and climate events, and the potential changes have a far greater impact on water resources, social and economic development than changes in the average climate. The results of long-term rainfall forecasts are an important reference for the monthly and seasonal production dispatching of the cascade dispatching and communication centers of water conservancy hubs, and are also important reference information for flood control dispatching and management departments to prepare flood control and power generation dispatching plans and reservoir water storage at the end of the flood season. The improvement of its accuracy can effectively improve the flood control, power generation, shipping, drought resistance, and ecological benefits of reservoirs, and realize the maximum utilization of water resources.

[0003] The existing long-term climate forecasts mainly focus on qualitative results based on trend forecast analysis, which is only a forecast description of the changing trend of average precipitation, temperature, etc. on the future monthly and seasonal time scales. This qualitative description is difficult to directly apply to predict future water inflow and cannot meet practical application needs.

[0004] The present invention proposes a regional drought and flood distribution prediction method and system based on big data and multiple models to solve the above-mentioned existing problems and improve the level of long-term quantitative rainfall forecasting. Summary of the invention

[0005] Purpose of the invention: To provide a regional drought and flood distribution prediction method based on big data and multiple models to solve the above problems existing in the prior art. On the other hand, to provide a regional drought and flood distribution prediction system based on big data and multiple models.

[0006] Technical solution: A regional drought and flood distribution prediction method based on big data and multiple models, including the following steps:

[0007] Step S1, collect historical rainfall and circulation sea temperature field data in the study area, screen historical samples of typical drought and flood years, calculate the SPI index of typical drought and flood years, and obtain the temporal and spatial characteristics of typical drought and flood years;

[0008] Step S2, constructing a conceptual physical model for typical drought and flood years and a forecast map of the regression cluster forecast physical model, and obtaining the abnormal characteristic value of the climate background and the predicted value of the drought and flood grade in the study area;

[0009] Step S3, constructing a precipitation field optimization prediction model based on singular value decomposition method;

[0010] Step S4: construct a drought and flood distribution prediction model for the study area based on the improved RegCM4 model to obtain the drought and flood distribution in the study area.

[0011] According to one aspect of the present application, step S1 further comprises:

[0012] Step S11, extracting the monthly average rainfall data of the study area, calculating the annual rainfall anomaly and flow anomaly in the main flood season, post-flood season, dry season and annual average, and selecting the candidate years and mandatory years of typical drought and flood years;

[0013] Step S12: based on the normal distribution curve of the SPI index, the drought and flood classification standard of the National Climate Center is used to classify the drought and flood grades, and the corresponding SPI index limit value is determined;

[0014] Step S13: Based on the SPI index of typical drought and flood years, the temporal and spatial characteristics of typical drought and flood years are obtained.

[0015] According to one aspect of the present application, step S13 is further:

[0016] Step S13a, counting the number of droughts and floods in the study area on a monthly, quarterly, and annual basis, and obtaining the drought and flood frequency of the study area;

[0017] Step S13b, using SPI at different time scales to represent different characteristics of drought and flood changes;

[0018] Step S13c: based on the spatial distribution of the linear change trend of the SPI index, obtain the spatial distribution of drought and flood trends in each season in the study area.

[0019] According to one aspect of the present application, step S2 further comprises:

[0020] Step S21, collecting climate factor data, screening out climate factor precursor signals based on the temporal and spatial characteristics of typical drought and flood years, constructing a conceptual physical model of typical drought and flood years, inputting climate factor precursor signals into the model, and calculating the abnormal characteristic values ​​of the climate background in the study area;

[0021] Step S22, collect historical hydrological forecasts and characteristic indexes of the study area, use the cause analysis method to construct a forecast equation, perform univariate regression forecasting, construct a regression cluster forecasting physical model forecast map, and obtain the predicted value of the drought and flood level in the study area.

[0022] According to one aspect of the present application, step S22 is further:

[0023] Step S22a, collecting historical hydrological forecasts and characteristic indexes of the study area, taking the characteristic indexes as forecasting factors, and constructing forecasting equations by solving regression equations, wherein the forecasting factors include: circulation, sea temperature, surface air pressure and air temperature;

[0024] Step S22b, sequentially perform the first univariate regression forecast on the circulation, sea temperature, surface pressure and air temperature with the study area, and obtain the univariate regression forecast value of the forecast factor;

[0025] Step S22c, the univariate regression forecast value of each forecast factor and the study area are subjected to a second multivariate regression quantitative forecast to obtain a positive contribution comprehensive index of each forecast factor, and a univariate regression forecast is performed to calculate the correlation coefficient of each forecast factor and the study area in turn to screen out the primary factors;

[0026] Step S22d, respectively calculate the positive contribution comprehensive forecast index of each preliminary factor and integrate the positive contribution comprehensive index of each forecast factor, perform univariate regression forecast, construct a regression cluster forecast physical model forecast map, and obtain the drought and flood grade prediction value of the study area.

[0027] According to one aspect of the present application, step S3 further comprises:

[0028] Step S31, collecting the previous climate elements closely related to the main flood season precipitation in the study area, and using the previous climate elements closely related to the main flood season precipitation in the study area as prediction factors;

[0029] Step S32: constructing a preliminary precipitation prediction model based on multiple climate factors using singular value decomposition method;

[0030] Step S33, constructing a prediction error correction model, coupling the precipitation field preliminary prediction model and the prediction error correction model to obtain a precipitation field optimization prediction model;

[0031] Step S34, input the time and spatial characteristics of typical drought and flood years into the precipitation field optimization prediction model, calculate the cumulative precipitation in the main flood season of the study area, the spatial distribution of precipitation anomalies in the main flood season, the location of the local abnormal precipitation center and the degree of precipitation anomaly.

[0032] According to one aspect of the present application, step S32 is further:

[0033] Step S32a, collecting precipitation data in the study area, global sea temperature data, Arctic sea ice data and Northern Hemisphere geopotential height data;

[0034] Step S32b, the precipitation data in the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere geopotential height data are processed for anomaly and SVD calculation is performed.

[0035] According to one aspect of the present application, step S33 is further:

[0036] Step S33a, screening the previous climate factor field that is closely related to precipitation in the study area;

[0037] Step S33b, divide the precipitation data of the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere potential height data into training data, verification data and test data in chronological order, use the training data to calibrate the parameters of the preliminary prediction model of the precipitation field, use the verification data to calibrate the parameters of the prediction error correction model, couple the preliminary prediction model of the precipitation field and the prediction error correction model to obtain the optimized prediction model of the precipitation field, and use the test data to calibrate the parameters of the optimized prediction model of the precipitation field.

[0038] According to one aspect of the present application, step S4 is further:

[0039] Step S41, improving the original boundary layer turbulence vertical transport scheme and cumulus parameterization scheme configuration of the RegCM4 model;

[0040] Step S42, screening out the optimal physical configuration combination scheme of the RegCM4 model;

[0041] Step S43, constructing a drought and flood distribution prediction model for the study area based on the improved RegCM4 model;

[0042] Step S44, input the abnormal characteristic value of the climate background in the study area, the predicted value of the drought and flood grade in the study area, the accumulated precipitation in the main flood season in the study area, the spatial distribution of precipitation anomaly in the main flood season, the location of the local abnormal precipitation center and the degree of precipitation anomaly into the drought and flood distribution prediction model of the study area to obtain the drought and flood distribution in the study area.

[0043] According to one aspect of the present application, step S41 is further:

[0044] Step S41a, improving the bidirectional synchronous coupling between the RegCM4 model and the ocean model, the input of sea temperature data and the configuration of the sea-air dynamic coupling;

[0045] Step S41b: Select the atmospheric and ocean driving fields as the boundary and initial fields of the RegCM4 model.

[0046] According to one aspect of the present application, step S42 is further:

[0047] Step S42a, selecting multiple physical schemes, performing precipitation simulation calculations on the configurations of the multiple physical schemes, and performing comparative evaluation and analysis, and screening out n physical configuration optimization combinations with better simulation capabilities, where n is a positive integer greater than 5;

[0048] Step S42b, using Monte Carlo and annealing chain methods to calibrate the thermal and dynamic parameters of the RegCM4 model, driving the regional climate model of the RegCM4 model to simulate the historical climate events of the study area, and evaluating the simulation capabilities of the regional climate model for the monthly, seasonal climate and anomalies of the study area under different physical configuration combinations;

[0049] Step S42c, using ocean model coupling and global climate model data to drive the RegCM4 model in turn, testing the reporting capabilities of different physical configuration combination models for monthly and seasonal climate anomalies in different regions of the study area, and screening out the optimal physical configuration combination scheme of the RegCM4 model.

[0050] According to another aspect of the present application, a regional drought and flood distribution prediction system based on big data and multiple models is provided, comprising:

[0051] at least one processor; and

[0052] a memory communicatively connected to at least one of the processors; wherein,

[0053] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the regional drought and flood distribution prediction method based on big data and multiple models described in any of the above technical solutions.

[0054] Beneficial effects: A regional drought and flood distribution prediction method based on big data and multiple models is adopted. Implicit information between data attributes is extracted through big data, and a drought and flood prediction model is constructed to provide an important method for long-term prediction. Big data technology is used to explore the potential relationship between climate influencing factor data, revealing strong climate factor signals or obvious characteristics that affect drought and flood events. At the same time, the study and application of climate dynamic models that calculate future weather conditions from the perspective of atmospheric dynamic movement realizes the transformation of long-term precipitation prediction from qualitative to quantitative, effectively improving the level of long-term quantitative rainfall forecasting, thereby better meeting practical application needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the present invention.

[0056] Figure 2 It is a flow chart of step S1 of the present invention.

[0057] Figure 3 It is a flow chart of step S2 of the present invention.

[0058] Figure 4 It is a flow chart of step S3 of the present invention.

[0059] Figure 5 It is a flow chart of step S4 of the present invention. DETAILED DESCRIPTION

[0060] like Figures 1 to 5 According to one aspect of the present application, a method for predicting regional drought and flood distribution based on big data and multiple models is provided, which is characterized by comprising the following steps:

[0061] Step S1, collect historical rainfall and circulation sea temperature field data in the study area, screen historical samples of typical drought and flood years, calculate the SPI index of typical drought and flood years, and obtain the temporal and spatial characteristics of typical drought and flood years;

[0062] Step S2, constructing a conceptual physical model for typical drought and flood years and a forecast map of the regression cluster forecast physical model, and obtaining the abnormal characteristic value of the climate background and the predicted value of the drought and flood grade in the study area;

[0063] Step S3, constructing a precipitation field optimization prediction model based on singular value decomposition method;

[0064] Step S4: construct a drought and flood distribution prediction model for the study area based on the improved RegCM4 model to obtain the drought and flood distribution in the study area.

[0065] According to one aspect of the present application, step S1 further comprises:

[0066] Step S11, extracting the monthly average rainfall data of the study area, calculating the annual rainfall anomaly and flow anomaly in the main flood season, post-flood season, dry season and annual average, and selecting the candidate years and mandatory years of typical drought and flood years;

[0067] In this embodiment, the monthly average rainfall data are used to calculate the annual main flood season, post-flood season, dry season and annual average rainfall deviations, and years with more than 10% more and less than 10% are selected as alternative years for typical drought and flood years. At the same time, the monthly average flow data are used to calculate the annual main flood season, post-flood season, dry season and annual average flow deviations, and years with more than 20% more and less than 20% are selected as alternative years for typical drought and flood years. Then, the five years before the maximum flow are selected as the mandatory years for typical flood years in the main flood season.

[0068] Based on the above-mentioned alternative years and mandatory years, the intersection of rainfall and flow is given priority, and then the principle of rainfall priority and the principle of priority for the absolute value of the anomaly are comprehensively weighed to select 10 typical drought and flood years in the main flood season and post-flood season.

[0069] Step S12: based on the normal distribution curve of the SPI index, the drought and flood classification standard of the National Climate Center is used to classify the drought and flood grades, and the corresponding SPI index limit value is determined;

[0070] Step S13: Based on the SPI index of typical drought and flood years, the temporal and spatial characteristics of typical drought and flood years are obtained.

[0071] The Standardized Precipitation Index, or SPI, is one of the indicators that characterizes the probability of precipitation in a certain period of time. For a certain region, the precipitation in a certain period of time generally fluctuates regularly. Based on this fact, if the precipitation in a certain period of time in a certain region is less than the multi-year average precipitation in that period of time to a certain extent, it is considered that a drought has occurred in that region during that period of time; conversely, if it is more than a certain extent, a flood will occur. Assuming that the change in precipitation follows the Gamma distribution, after calculating the cumulative frequency distribution of precipitation in a certain period of time, the cumulative frequency distribution of precipitation is converted into a standard normal distribution using mathematical methods, and finally the corresponding SPI index is obtained.

[0072] The SPI index is dimensionless and standardized. It can compare precipitation levels in different regions and at different times, and can better reflect the intensity and duration of drought. The SPI index has the advantage of multiple time scales. It can not only reflect changes in precipitation in short time scales, but also reflect the evolution of water resources in long time scales.

[0073] Assuming the precipitation at a certain time scale is x, the probability density function of its Gamma distribution is:

[0074]

[0075]

[0076] Where: and are shape parameter and scale parameter respectively; is the gamma function;

[0077] Using the maximum likelihood method to estimate and Value, that is:

[0078]

[0079]

[0080]

[0081] In the formula, x is the precipitation data sample, is the climate mean of precipitation, n is the length of the calculation sequence;

[0082] Since the gamma function does not include the case of x=0, and the actual precipitation can be 0, It should be the mean of the non-zero items in the precipitation series. If the length of the precipitation series is n, and the number of zero items is m, let , then the cumulative probability of a certain time scale is as follows:

[0083]

[0084] Where: , that is .

[0085] Then the cumulative probability distribution The corresponding SPI value can be obtained by converting it into a standard normal distribution. Since the SPI calculates the drought index by calculating the cumulative probability of precipitation at a given time scale, the total precipitation of the current month and the previous i months (i = 1, 2, 3, ...) is used to calculate the SPI with a time scale of i+1 month.

[0086] According to one aspect of the present application, step S13 is further:

[0087] Step S13a, counting the number of droughts and floods in the study area on a monthly, quarterly, and annual basis, and obtaining the drought and flood frequency of the study area;

[0088] Step S13b, using SPI at different time scales to represent different characteristics of drought and flood changes;

[0089] Step S13c, based on the spatial distribution of the linear change trend of the SPI index, the spatial distribution of drought and flood trends in each season of the study area is obtained. In this embodiment, based on the standardized precipitation index SPI, a linear trend analysis method is used for analysis and significance test, and three evaluation indicators characterizing regional drought and flood characteristics are introduced in the analysis, namely, drought and flood frequency, drought and flood station ratio, and drought and flood intensity. The evaluation indicators are described as follows:

[0090] Drought (flood) frequency, Drought (flood) frequency is the number of years of drought (flood) at station i in the study area The ratio of drought (flood) to the total number of years N, that is, the frequency of drought (flood) occurrence, is calculated as follows:

[0091] Drought / flood station ratio: Drought / flood station ratio is the number of droughts (floods) that occurred in the study area in year j The proportion of the total number of stations M is calculated as:

[0092] The classification criteria are: ≥50% means drought (flood) in the whole region; 50%> ≥33% is regional drought (flood); 33%> ≥25% means drought (flood) in some areas; 25%> ≥10% means drought (flood) in some areas; <10% means no obvious drought (flood)

[0093] Drought (flood) intensity: Drought (flood) intensity is the average absolute value of SPI of station i where drought (flood) occurs in year j, and the calculation formula is:

[0094]

[0095] Where: m is the number of stations where drought or flood events occurred; i is the station; j is the year; is the absolute value of SPI for site i in year j;

[0096] Regional drought and flood calculation: The method for calculating drought and flood in a certain area is to use the SPI values ​​or drought and flood intensity index of all stations in the area for average calculation, and the value obtained is the corresponding drought and flood index of the area.

[0097] According to one aspect of the present application, step S2 further comprises:

[0098] Step S21, collecting climate factor data, screening out climate factor precursor signals based on the temporal and spatial characteristics of typical drought and flood years, constructing a conceptual physical model of typical drought and flood years, inputting climate factor precursor signals into the model, and calculating the abnormal characteristic values ​​of the climate background in the study area;

[0099] Step S22, collect historical hydrological forecasts and characteristic indexes of the study area, use the cause analysis method to construct a forecast equation, perform univariate regression forecasting, construct a regression cluster forecasting physical model forecast map, and obtain the predicted value of the drought and flood level in the study area.

[0100] The big data used in drought and flood predictions include meteorological data, sea temperature data, sea ice and snow data, land surface monitoring data, astronomical data, reanalysis data, numerical forecast data and other data.

[0101] In this embodiment, a research approach combining theoretical research, method analysis and experiment is adopted to systematically analyze the characterization patterns of long-term rainfall in the upper reaches of the Yangtze River in multi-source heterogeneous data such as different meteorological data, sea temperature data, sea ice data, snow data, land surface monitoring data, astronomical data, reanalysis data, and numerical forecast data, sort out the data types associated with the long-term rainfall change process in the study area, solve the problem of extracting the change process data sequence from the multi-source heterogeneous data and judging its correlation with the rainfall change process, perform correlation calculation for big data analysis based on various types of multi-source heterogeneous data, analyze the correlation between the multi-source heterogeneous data in the study area and the long-term rainfall process through correlation analysis, organize the data cluster sequence according to the size of the correlation according to the structured and unstructured data types, and then sort out the degree of influence of different influencing factors on rainfall, and construct a multi-temporal and multi-factor long-term rainfall forecasting method suitable for big data analysis.

[0102] According to one aspect of the present application, step S22 is further:

[0103] Step S22a, collecting historical hydrological forecasts and characteristic indexes of the study area, taking the characteristic indexes as forecasting factors, and constructing forecasting equations by solving regression equations, wherein the forecasting factors include: circulation, sea temperature, surface air pressure and air temperature;

[0104] Step S22b, sequentially perform the first univariate regression forecast on the circulation, sea temperature, surface pressure and air temperature with the study area, and obtain the univariate regression forecast value of the forecast factor;

[0105] Step S22c, the univariate regression forecast value of each forecast factor and the study area are subjected to a second multivariate regression quantitative forecast to obtain a positive contribution comprehensive index of each forecast factor, and a univariate regression forecast is performed to calculate the correlation coefficient of each forecast factor and the study area in turn to screen out the primary factors;

[0106] Step S22d, respectively calculate the positive contribution comprehensive forecast index of each preliminary factor and integrate the positive contribution comprehensive index of each forecast factor, perform univariate regression forecast, construct a regression cluster forecast physical model forecast map, and obtain the drought and flood grade prediction value of the study area.

[0107] A conceptual model is a more or less formal description, a simplified description of reality, including a certain degree of abstraction, explicitly or implicitly constructed according to the exact way of use in the mind, and describes and explains the laws and mechanisms of things in the form of flowcharts composed of diagrams, text, symbols, etc. The characteristics of a conceptual model are that the diagrams are relatively intuitive and patterned, and the text is connected by symbols such as arrows. The keywords are relatively concise and clear. They can reveal the main characteristics and essence of things, and are intuitive and easy to understand. The physical model is based on the natural things that can be seen by oneself. The conversion of the conceptual model is based on what is understood into a rare language. Since the actual physical phenomena and physical laws are generally very complex and involve many factors, the minor factors are discarded, the main factors are grasped, and the essential characteristics of objective things are highlighted. The conceptual model established is called a physical model. The conceptual physical model often converts the complex actual situation into a certain simple situation that is easy to accept, thereby forming a certain empirical law.

[0108] The conceptual physical model analysis method of drought and flood is to select the main common factors based on the study of various relevant "big" data in typical historical years. With the rapid development of meteorological monitoring technology and data sharing system, a three-dimensional meteorological element multi-temporal and spatial scale monitoring network from the sky to the ground has been realized, providing a data basis for the application of big data analysis technology. It is based on data science, follows the rules of the fourth paradigm of scientific research, and is data-centric. It analyzes complex and massive data from the perspective of "all non-sampling, efficiency non-precision, and correlation non-causality" to discover the correlation and evolution characteristics between things, find out the factors and factor states related to the forecast object, and determine the common factors.

[0109] The cluster analysis method is a multi-factor comprehensive forecasting method. It analyzes the statistical correlation between the forecast elements and the previous multiple factors, integrates them with mathematical statistics, and makes forecasts. It mainly uses correlation and multivariate regression analysis.

[0110] The cluster analysis method first calculates the correlation coefficient and related probability between the forecast object and the above-mentioned structured data month by month and grid by grid. The sample time of the census is from January of the previous year to February of the current year of the historical series. In the census results, the grids with correlation coefficients that meet the standards are screened out as forecast factors. After multiple processing, univariate regression is established with the positive contribution comprehensive index of each screened forecast factor. The univariate regression value avoids the influence of the extreme value of individual forecast factors on the comprehensive forecast opinion; the quadratic comprehensive regression fitting forecast value is obtained by calculating the regression forecast value calculated with the original values ​​of multiple forecast factors. The univariate regression value and the quadratic comprehensive regression fitting forecast value are used as the vertical axis and the horizontal axis respectively to make the forecast physical model.

[0111] According to one aspect of the present application, step S3 further comprises:

[0112] Step S31, collecting the previous climate elements closely related to the main flood season precipitation in the study area, and using the previous climate elements closely related to the main flood season precipitation in the study area as prediction factors;

[0113] Step S32: constructing a preliminary precipitation prediction model based on multiple climate factors using singular value decomposition method;

[0114] Step S33, constructing a prediction error correction model, coupling the precipitation field preliminary prediction model and the prediction error correction model to obtain a precipitation field optimization prediction model;

[0115] Step S34, input the time and spatial characteristics of typical drought and flood years into the precipitation field optimization prediction model, calculate the cumulative precipitation in the main flood season of the study area, the spatial distribution of precipitation anomalies in the main flood season, the location of the local abnormal precipitation center and the degree of precipitation anomaly.

[0116] Singular value decomposition is a simple and practical analysis technique that discovers interesting associations and correlations between item sets from a large amount of data. It is used to discover the associations or correlations that exist in a large amount of data sets, thereby describing the laws and patterns of the simultaneous occurrence of certain attributes in a thing. It is specifically based on the maximum covariance between two element fields. It is an effective diagnostic analysis method for studying the correlation structure of two fields. It is suitable for climate diagnostic analysis and teleconnection of large-scale meteorological fields. The singular value decomposition method can separate the independent coupled distribution patterns and the orthogonality that each satisfies to the greatest extent, reveal the spatial connection of the time domain correlation between the two meteorological fields, and find the key areas of teleconnection patterns and mutual connections.

[0117] In this embodiment, the precipitation field optimization prediction model based on the singular value decomposition method has an overall better prediction effect than the prediction error correction model. The systematic prediction error of the traditional singular value decomposition method can be eliminated to a certain extent through secondary correction. At the same time, it can effectively predict the cumulative precipitation in the main flood season and can basically predict the spatial distribution of precipitation anomalies in the main flood season.

[0118] In general, compared with the traditional singular value decomposition method, the precipitation field optimization prediction model using multiple climate factors and quadratic correction has certain advantages in predicting drought and flood distribution in the study area.

[0119] According to one aspect of the present application, step S32 is further:

[0120] Step S32a, collecting precipitation data in the study area, global sea temperature data, Arctic sea ice data and Northern Hemisphere geopotential height data;

[0121] Step S32b, the precipitation data in the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere geopotential height data are processed for anomaly and SVD calculation is performed.

[0122] According to one aspect of the present application, step S33 is further:

[0123] Step S33a, screening the previous climate factor field that is closely related to precipitation in the study area;

[0124] Step S33b, divide the precipitation data of the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere potential height data into training data, verification data and test data in chronological order, use the training data to calibrate the parameters of the preliminary prediction model of the precipitation field, use the verification data to calibrate the parameters of the prediction error correction model, couple the preliminary prediction model of the precipitation field and the prediction error correction model to obtain the optimized prediction model of the precipitation field, and use the test data to calibrate the parameters of the optimized prediction model of the precipitation field.

[0125] During the training period, the data needs to be reconstructed into a matrix, converting the left-field precipitation grid data into a time matrix of X(m,k), and the right-field climate factor grid data into a time matrix of Y(n,k), where m and n are the sum of the number of precipitation data grid points and the number of selected climate factor data grid points, respectively, and k is the duration of the training period; during the verification period, the observed precipitation data and the SVDF model prediction results are reconstructed in the same matrix;

[0126] The anomaly sign consistency rate is used to quantitatively evaluate the prediction accuracy of precipitation anomaly conditions.

[0127] According to one aspect of the present application, step S4 is further:

[0128] Step S41, improving the original boundary layer turbulence vertical transport scheme and cumulus parameterization scheme configuration of the RegCM4 model;

[0129] Step S42, screening out the optimal physical configuration combination scheme of the RegCM4 model;

[0130] Step S43, constructing a drought and flood distribution prediction model for the study area based on the improved RegCM4 model;

[0131] Step S44, input the abnormal characteristic value of the climate background in the study area, the predicted value of the drought and flood grade in the study area, the accumulated precipitation in the main flood season in the study area, the spatial distribution of precipitation anomaly in the main flood season, the location of the local abnormal precipitation center and the degree of precipitation anomaly into the drought and flood distribution prediction model of the study area to obtain the drought and flood distribution in the study area.

[0132] According to one aspect of the present application, step S41 is further:

[0133] Step S41a, improving the bidirectional synchronous coupling between the RegCM4 model and the ocean model, the input of sea temperature data and the configuration of the sea-air dynamic coupling;

[0134] Step S41b: Select the atmospheric and ocean driving fields as the boundary and initial fields of the RegCM4 model.

[0135] According to one aspect of the present application, step S42 is further:

[0136] Step S42a, selecting multiple physical schemes, performing precipitation simulation calculations on the configurations of the multiple physical schemes, and performing comparative evaluation and analysis, and screening out n physical configuration optimization combinations with better simulation capabilities, where n is a positive integer greater than 5;

[0137] Step S42b, using Monte Carlo and annealing chain methods to calibrate the thermal and dynamic parameters of the RegCM4 model, driving the regional climate model of the RegCM4 model to simulate the historical climate events of the study area, and evaluating the simulation capabilities of the regional climate model for the monthly, seasonal climate and anomalies of the study area under different physical configuration combinations;

[0138] Step S42c, using ocean model coupling and global climate model data to drive the RegCM4 model in turn, testing the reporting capabilities of different physical configuration combination models for monthly and seasonal climate anomalies in different regions of the study area, and screening out the optimal physical configuration combination scheme of the RegCM4 model.

[0139] According to another aspect of the present application, a regional drought and flood distribution prediction system based on big data and multiple models is provided, characterized in that it includes:

[0140] at least one processor; and

[0141] a memory communicatively connected to at least one of the processors; wherein,

[0142] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement any of the above-mentioned regional drought and flood distribution prediction methods based on big data and multiple models.

[0143] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A regional drought and flood distribution prediction method based on big data and multiple models, characterized by: The steps include: Step S1, collect historical rainfall and circulation sea temperature field data in the study area, screen historical samples of typical drought and flood years, calculate the SPI index of typical drought and flood years, and obtain the temporal and spatial characteristics of typical drought and flood years; Step S2, constructing a conceptual physical model for typical drought and flood years and a forecast map of the regression cluster forecast physical model, and obtaining the abnormal characteristic value of the climate background and the predicted value of the drought and flood grade in the study area; Step S3, constructing a precipitation field optimization prediction model based on singular value decomposition method; Step S4: construct a drought and flood distribution prediction model for the study area based on the improved RegCM4 model to obtain the drought and flood distribution in the study area.

2. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 1, characterized in that: The step S1 is further as follows: Step S11, extracting the monthly average rainfall data of the study area, calculating the annual rainfall anomaly and flow anomaly in the main flood season, post-flood season, dry season and annual average, and selecting the candidate years and mandatory years of typical drought and flood years; Step S12: based on the normal distribution curve of the SPI index, the drought and flood classification standard of the National Climate Center is used to classify the drought and flood grades, and the corresponding SPI index limit value is determined; Step S13, based on the SPI index of typical drought and flood years, obtaining the temporal and spatial characteristics of typical drought and flood years; The step S13 is further as follows: Step S13a, counting the number of droughts and floods in the study area on a monthly, quarterly, and annual basis, and obtaining the drought and flood frequency of the study area; Step S13b, using SPI at different time scales to represent different characteristics of drought and flood changes; Step S13c: based on the spatial distribution of the linear change trend of the SPI index, obtain the spatial distribution of drought and flood trends in each season in the study area.

3. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 1, characterized in that: The step S2 is further as follows: Step S21, collecting climate factor data, screening out climate factor precursor signals based on the temporal and spatial characteristics of typical drought and flood years, constructing a conceptual physical model of typical drought and flood years, inputting climate factor precursor signals into the model, and calculating the abnormal characteristic values ​​of the climate background in the study area; Step S22, collect historical hydrological forecasts and characteristic indexes of the study area, use the cause analysis method to construct a forecast equation, perform univariate regression forecasting, construct a regression cluster forecasting physical model forecast map, and obtain the predicted value of the drought and flood level in the study area.

4. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 3 is characterized in that: The step S22 is further as follows: Step S22a, collecting historical hydrological forecasts and characteristic indexes of the study area, taking the characteristic indexes as forecasting factors, and constructing forecasting equations by solving regression equations, wherein the forecasting factors include: circulation, sea temperature, surface air pressure and air temperature; Step S22b, sequentially perform the first univariate regression forecast on the circulation, sea temperature, surface pressure and air temperature with the study area, and obtain the univariate regression forecast value of the forecast factor; Step S22c, the univariate regression forecast value of each forecast factor and the study area are subjected to a second multivariate regression quantitative forecast to obtain a positive contribution comprehensive index of each forecast factor, and a univariate regression forecast is performed to calculate the correlation coefficient of each forecast factor and the study area in turn to screen out the primary factors; Step S22d, respectively calculate the positive contribution comprehensive forecast index of each preliminary factor and integrate the positive contribution comprehensive index of each forecast factor, perform univariate regression forecast, construct a regression cluster forecast physical model forecast map, and obtain the drought and flood grade prediction value of the study area.

5. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 1, characterized in that: The step S3 is further as follows: Step S31, collecting the previous climate elements closely related to the main flood season precipitation in the study area, and using the previous climate elements closely related to the main flood season precipitation in the study area as prediction factors; Step S32: constructing a preliminary precipitation prediction model based on multiple climate factors using singular value decomposition method; Step S33, constructing a prediction error correction model, coupling the precipitation field preliminary prediction model and the prediction error correction model to obtain a precipitation field optimization prediction model; Step S34, input the time and spatial characteristics of typical drought and flood years into the precipitation field optimization prediction model, calculate the cumulative precipitation in the main flood season of the study area, the spatial distribution of precipitation anomalies in the main flood season, the location of the local abnormal precipitation center and the degree of precipitation anomaly.

6. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 5 is characterized in that: The step S32 is further as follows: Step S32a, collecting precipitation data in the study area, global sea temperature data, Arctic sea ice data and Northern Hemisphere geopotential height data; Step S32b, the precipitation data in the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere geopotential height data are processed for anomaly and SVD calculation is performed.

7. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 5, characterized in that: The step S33 is further as follows: Step S33a, screening the previous climate factor field that is closely related to the precipitation in the study area; Step S33b, divide the precipitation data of the study area, the global sea temperature data, the Arctic sea ice data and the Northern Hemisphere potential height data into training data, verification data and test data in chronological order, use the training data to calibrate the parameters of the preliminary prediction model of the precipitation field, use the verification data to calibrate the parameters of the prediction error correction model, couple the preliminary prediction model of the precipitation field and the prediction error correction model to obtain the optimized prediction model of the precipitation field, and use the test data to calibrate the parameters of the optimized prediction model of the precipitation field.

8. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 1, characterized in that: The step S4 is further as follows: Step S41, improving the original boundary layer turbulence vertical transport scheme and cumulus parameterization scheme configuration of the RegCM4 model; Step S42, screening out the optimal physical configuration combination scheme of the RegCM4 model; Step S43, constructing a drought and flood distribution prediction model for the study area based on the improved RegCM4 model; Step S44, input the abnormal characteristic value of the climate background in the study area, the predicted value of the drought and flood grade in the study area, the accumulated precipitation in the main flood season in the study area, the spatial distribution of the abnormal precipitation in the main flood season, the location of the local abnormal precipitation center and the degree of abnormal precipitation into the drought and flood distribution prediction model of the study area to obtain the drought and flood distribution in the study area; The step S41 is further as follows: Step S41a, improving the bidirectional synchronous coupling between the RegCM4 model and the ocean model, the input of sea temperature data and the configuration of the sea-air dynamic coupling; Step S41b: Select the atmospheric and ocean driving fields as the boundary and initial fields of the RegCM4 model.

9. The regional drought and flood distribution prediction method based on big data and multiple models as claimed in claim 8, characterized in that: The step S42 is further as follows: Step S42a, selecting multiple physical schemes, performing precipitation simulation calculations on the configurations of the multiple physical schemes, and performing comparative evaluation and analysis, and screening out n physical configuration optimization combinations with better simulation capabilities, where n is a positive integer greater than 5; Step S42b, using Monte Carlo and annealing chain methods to calibrate the thermal and dynamic parameters of the RegCM4 model, driving the regional climate model of the RegCM4 model to simulate the historical climate events of the study area, and evaluating the simulation capabilities of the regional climate model for the monthly, seasonal climate and anomalies of the study area under different physical configuration combinations; Step S42c, using ocean model coupling and global climate model data to drive the RegCM4 model in turn, testing the reporting capabilities of different physical configuration combination models for monthly and seasonal climate anomalies in different regions of the study area, and screening out the optimal physical configuration combination scheme of the RegCM4 model.

10. Regional drought and flood distribution prediction system based on big data and multiple models, characterized by: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the regional drought and flood distribution prediction method based on big data and multiple models as described in any one of claims 1 to 9.

Images