CN108761574B - Rainfall estimation method based on multi-source information fusion - Google Patents
Rainfall estimation method based on multi-source information fusion Download PDFInfo
- Publication number
- CN108761574B CN108761574B CN201810427354.2A CN201810427354A CN108761574B CN 108761574 B CN108761574 B CN 108761574B CN 201810427354 A CN201810427354 A CN 201810427354A CN 108761574 B CN108761574 B CN 108761574B
- Authority
- CN
- China
- Prior art keywords
- rainfall
- data
- satellite
- source
- research area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/14—Rainfall or precipitation gauges
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
Abstract
The invention provides a rainfall estimation method based on multi-source information fusion, which comprises the following steps: combining the rainfall data observed by the ground station in the research area with the rainfall data of the multisource satellite in the research area to form a multisource data set of the research area; establishing a Bayesian rainfall prediction model of a research area based on a dynamic Bayesian theory; solving a nonlinear optimal solution of a Bayesian rainfall prediction model by using a maximum entropy method, and further determining optimal weight and uncertainty information of each satellite data source; and generating an estimation result of the rainfall amount of the multi-source information fusion applied in the research area based on the optimal weight and uncertainty of the satellite data source. Has the advantages that: according to the method, the result of fusion analysis of multiple data sources reduces the uncertainty of regional rainfall estimation caused by the inaccuracy of a single rainfall information; the method provides more reliable data input and richer and more refined modeling data for strengthening regional high-precision disaster early warning, avoiding flood risks or estimating torrential rain and flood in a small watershed.
Description
Technical Field
The invention belongs to the technical field of rainfall estimation, and particularly relates to a rainfall estimation method based on multi-source information fusion.
Background
Rainfall is a key link in the circulation process of a hydrological system, plays a vital role in the whole water circulation, and the huge change of rainfall in a short time easily causes regional geological and environmental problems, such as urban ponding, mountain geological disasters or river basin flood disasters, and the like, and causes serious damage to the development of social politics and economy. How to accurately estimate the regional rainfall in a short time is one of the key problems to be solved urgently in the field of water information at present.
With the development of observation means and level, the technology for observing the ground rainfall by the satellite makes great progress, and provides a data base for estimating the ground rainfall. Each type of satellite observation rainfall data originates from different platforms and has respective formats and standards; at present, rainfall estimation is carried out by independently using one type of data, so that deviation of estimation results along with errors of observation data is easily caused, and accurate data support is not provided for geological disaster monitoring, flood prevention, drought resistance and other works.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a rainfall estimation method based on multi-source information fusion, which can effectively solve the problems.
The technical scheme adopted by the invention is as follows:
the invention provides a rainfall estimation method based on multi-source information fusion, which comprises the following steps:
step 1, acquiring multi-source satellite rainfall data and ground station rainfall observation data in a research area;
step 3, combining the rainfall data observed by the ground station in the research area with the rainfall data of the multisource satellite in the research area to form a multisource data set of the research area;
step 5, extracting a target dynamic training sample: aiming at the Bayes rainfall prediction model established in the step 4, training the Bayes rainfall prediction model established in the step 4 by using the rainfall data observed by the ground station as a training sample to obtain a trained Bayes rainfall prediction model;
and 7, generating an estimation result of the rainfall amount fused by the multi-source information in the research area based on the optimal weight and uncertainty of the satellite data source.
Preferably, the multi-source satellite rainfall data refers to rainfall data from different platforms and different satellite types; after original multi-source satellite rainfall data are obtained, abnormal values are removed, and then format unification, area cutting and scale matching processing are carried out.
Preferably, in step 2, the specific process of format unification is as follows: unifying the formats of the multi-source satellite rainfall data into a binary format; the header file comprises a coordinate range and a data source, and the file body comprises a rainfall data matrix at a corresponding position;
the specific process of the region cutting is as follows: cutting out multi-source satellite rainfall data belonging to a research area range from the multi-source satellite rainfall data, wherein the space range of the cut-out data is a rectangle, and the coordinate range of the cut-out data is determined by an extreme value of a research area coordinate, namely the cut-out area is a minimum rectangle containing the research area;
the specific process of the scale matching is as follows: the scale matching process comprises time unification and space unification; because the monitoring time and the spatial resolution capability of various satellite rainfall data are different, the multisource satellite rainfall data in the research area range are converted into data with uniform time and uniform spatial grids;
and converting the multi-source satellite rainfall data into a data body with uniform format, consistent time interval and space position through format unification, region cutting and scale matching, thereby forming a multi-source satellite rainfall data body of rainfall at each moment in a standard grid in the research region.
Preferably, step 3 specifically comprises:
putting the rainfall data observed by the ground stations into corresponding grids of a multisource satellite rainfall data volume in a research area, and taking an average value if a plurality of ground stations of the same ground station exist in a single grid for observing the rainfall data; taking grids with ground station rainfall observation data as reference points and constraint conditions, and evaluating and optimizing the conformity of the optimal solution of the Bayesian rainfall prediction model; and for grids without the data of the rainfall observed by the ground station, estimating by using an optimized Bayesian rainfall prediction model to obtain an optimal solution.
Preferably, in step 4, the established bayesian rainfall prediction model is:
wherein p (y | D) is the posterior probability of the fused target rainfall;
y is fused target precipitation;
d is ground observation rainfall data in a certain period of time;
p(fki D) is the posterior probability of the rainfall inverted by the data under the condition of ground rainfall observation, and is also regarded as the accuracy of the rainfall inverted by different satellites under the condition of ground rainfall observation;
fkprecipitation for inversion of different models;
p(y|fkd) inverting precipitation based on different models, and observing the posterior distribution of the fusion precipitation under the precipitation condition on the ground;
k is different model types;
m is the total amount of the model;
the mean value thereof is expressed as
Wherein:
e (y | D) represents the mean of the Bayesian multi-model ensemble, wherein the E symbol represents the mean;
E[pk(y|fk,D)]the mean value of the posterior distribution of the fusion precipitation under the condition of the ground observation precipitation is used for inverting the precipitation based on different models;
wkthe shorthand of the posterior probability of rainfall is inverted for data under the condition of ground observation rainfall;
wk＝p(fki D), then have
The variance is expressed as:
wherein:
var (y | D) represents the variance based on dynamic Bayesian multi-model integration;
σ2is the variance of satellite precipitation relative to ground precipitation.
Preferably, in step 6, when the maximum entropy method is used to solve the nonlinear optimal solution of the trained bayesian rainfall prediction model, the corresponding conditional probability when the conditional entropy is maximum is the conditional probability that needs to be obtained, that is:
wherein:
h (p) is the conditional entropy when the fusion rainfall and the observation correlation of the ground station are maximum;
and p (D) is the fusion of the rainfall amount as the ground observation value.
Preferably, step 7 specifically comprises:
interpolating the optimal weight and uncertainty information of a fixed station in a target period to the whole research area by using a geostatistical spatial interpolation method, and carrying out normalization processing on the interpolated weight to ensure that the sum of the weights of each satellite of each grid point is always equal to 1, thereby obtaining the normalized weight of various satellite data in the research area and obtaining a weight spatial distribution map of the research area; and generating an estimation result of applying multi-source information fusion rainfall in the research area based on the weight space distribution map.
The rainfall estimation method based on multi-source information fusion provided by the invention has the following advantages:
the method selects suitable multisource multi-platform satellite rainfall monitoring data in a target area, establishes a rainfall model based on a dynamic Bayesian model, and obtains nonlinear extreme points of the model by an entropy maximization method, so as to obtain global optimal weight and uncertainty of various data, and finally estimates the rainfall in continuous time periods of the target area. The rainfall estimation method based on multi-source information fusion can exert the advantages of multi-source data and obtain rainfall data with higher precision and higher confidence. According to the method, the result of fusion analysis of multiple data sources reduces the uncertainty of regional rainfall estimation caused by the inaccuracy of a single rainfall information; the method provides more reliable data input and richer and more refined modeling data for strengthening regional high-precision disaster early warning, avoiding flood risks or estimating torrential rain and flood in a small watershed.
Drawings
FIG. 1 is a schematic flow chart of a rainfall estimation method based on multi-source information fusion according to the present invention;
FIG. 2 is a comparison graph of the statistical rules of the multi-source information estimation results and the estimation results of a single satellite according to the present invention;
FIG. 3 is a comparison graph of rainfall estimated proximity value and rainfall for single satellite data based on multi-source information of the present invention;
FIG. 4 is a plot of the estimated root mean square error of rainfall based on multi-source information of the present invention versus rainfall for single satellite data.
Detailed Description
In order to make the technical problems, technical solutions and advantageous effects solved by the present invention more clearly apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The comprehensive utilization of multi-platform data for rainfall analysis and estimation is a great trend. The rainfall of the multi-source satellite from multiple platforms is fused, so that the advantages of data of each platform can be fully exerted, and estimation result errors caused by single-type data deviation can be reduced. Meanwhile, the rainfall amount is corrected by taking the ground monitoring rainfall amount as a standard, a set of multi-source multi-information fusion rainfall amount estimation method can be established, more reliable and accurate ground rainfall amount distribution is provided, and accurate data support is further provided for geological disaster monitoring, flood prevention, drought resistance and other works.
The invention relates to a multi-source information fusion rainfall estimation method, which utilizes multi-source multi-platform rainfall monitoring information to form rainfall spatial distribution data; the rainfall characteristics are analyzed by establishing a Bayesian model suitable for multi-source data, and the rainfall at a certain moment is estimated by combining historical rainfall information.
According to the method, the most effective rainfall data in the area is extracted through the arrangement and analysis of the multi-source rainfall information; and integrating and preprocessing different types of data with different precisions to obtain a multi-information data body at the same position in the region. The method is based on Bayesian theory to fuse and analyze the data volume to obtain rainfall estimation information of the area, so that the effective part of multi-source information is mined to the maximum extent and the influence of single-class data deviation is reduced. And carrying out optimization solution on the dynamic Bayesian model by an entropy maximum method. The result of fusion analysis of multiple data sources reduces the uncertainty of regional rainfall estimation caused by the inaccuracy of a single rainfall information; the method provides more reliable data input and richer and more refined modeling data for strengthening regional high-precision disaster early warning, avoiding flood risks or estimating torrential rain and flood in a small watershed.
The invention provides a rainfall estimation method based on multi-source information, which mainly comprises the following design concepts: firstly, utilizing multisource multi-platform satellite rainfall data and ground station rainfall monitoring data, and carrying out preprocessing processes such as format unification, region cutting, scale matching and the like on the satellite data to form a multisource data body of a research region; secondly, combining rainfall data monitored by using a ground station with multi-source rainfall data to establish a multi-source multi-platform rainfall information data set; thirdly, establishing a rainfall prediction model of the research area based on a dynamic Bayesian theory, and comprehensively reflecting the influence of various data on rainfall; solving a nonlinear optimal solution for establishing the model by using a maximum entropy method to obtain the estimation weight and uncertainty of various data; and fifthly, generating a result of applying multi-source information fusion rainfall estimation in the research area.
Specifically, referring to fig. 1, the rainfall estimation method based on multi-source information fusion provided by the invention utilizes multi-source multi-platform satellite rainfall data and ground station monitoring rainfall data information, and forms a multi-source data body of a research area by performing preprocessing processes such as format unification, area clipping, scale matching and the like on the satellite data; then, establishing a multi-source multi-platform rainfall information data set by using rainfall data monitored by the ground station; then, establishing a precipitation prediction model of the research area based on a dynamic Bayesian theory, and comprehensively reflecting the influence of various data on precipitation; and then, solving a nonlinear optimal solution for establishing the model by using a maximum entropy method, obtaining the calculation weight of various data, and further generating the result and uncertainty of the precipitation estimation of the multi-source information applied in the research area. The method comprises the following specific steps: (1) unifying the format of multi-source information of a research area, cutting the area and matching the scale; (2) establishing a multi-source multi-platform rainfall information data set of a research area; (3) establishing a dynamic Bayesian model by adopting the data set; (4) adopting a maximum entropy algorithm nonlinear optimization solution of the dynamic Bayesian model; (5) and calculating the estimated value of the precipitation of the multi-source information by adopting the solving result.
The method comprises the following steps:
step 1, acquiring multi-source satellite rainfall data and ground station rainfall observation data in a research area;
the multi-source satellite rainfall data refers to rainfall data from different platforms and different satellite types; after original multi-source satellite rainfall data are obtained, abnormal values are removed, and then format unification, area cutting and scale matching processing are carried out.
in step 2, the specific process of format unification is as follows: unifying the formats of the multi-source satellite rainfall data into a binary format; the header file comprises a coordinate range and a data source, and the file body comprises a rainfall data matrix at a corresponding position;
the specific process of the region cutting is as follows: cutting out multi-source satellite rainfall data belonging to a research area range from the multi-source satellite rainfall data, wherein the space range of the cut-out data is a rectangle, and the coordinate range of the cut-out data is determined by an extreme value of a research area coordinate, namely the cut-out area is a minimum rectangle containing the research area;
the specific process of the scale matching is as follows: the scale matching process comprises time unification and space unification; because the monitoring time and the spatial resolution capability of various satellite rainfall data are different, the multisource satellite rainfall data in the research area range are converted into data with uniform time and uniform spatial grids;
and converting the multi-source satellite rainfall data into a data body with uniform format, consistent time interval and space position through format unification, region cutting and scale matching, thereby forming a multi-source satellite rainfall data body of rainfall at each moment in a standard grid in the research region.
By the steps, multi-source precipitation data are comprehensively utilized, and satellite precipitation data of different types of different platforms, such as various data in formats of TRMM 3B42RT, TRMM 3B42V7, GPM IMERG, PERSIANN, GsMap, CPC CMORPH, FY-2E and the like, are preprocessed. The main work on the basis of removing abnormal values comprises format unification, area clipping and scale unification. The formats of different satellite precipitation data are converted into uniform formats through format unification, so that further modeling calculation is facilitated; the region cutting is used for intercepting a research region from a global data volume; and the uniform scale unifies the spatial resolution of various data, and lays a foundation for modeling. The form of the multi-source satellite data is unified after the process.
Regularizing the preprocessed satellite data, and converting the multi-source satellite data into a data body with consistent time interval and spatial position, namely establishing the multi-source data body of rainfall at each moment in a standard grid in a research area. And combining the rainfall observed on the ground with the multi-source satellite data to form a multi-source multi-platform rainfall information data set of the research area. The ground rainfall at the corresponding position is an important practical basis for evaluating the accuracy of various satellite data.
Step 3, combining the rainfall data observed by the ground station in the research area with the rainfall data of the multisource satellite in the research area to form a multisource data set of the research area;
the step 3 specifically comprises the following steps:
putting the rainfall data observed by the ground stations into corresponding grids of a multisource satellite rainfall data volume in a research area, and taking an average value if a plurality of ground stations of the same ground station exist in a single grid for observing the rainfall data; taking grids with ground station rainfall observation data as reference points and constraint conditions, and evaluating and optimizing the conformity of the optimal solution of the Bayesian rainfall prediction model; and for grids without the data of the rainfall observed by the ground station, estimating by using an optimized Bayesian rainfall prediction model to obtain an optimal solution.
in the step, a dynamic Bayesian model is established according to the established precipitation information data set, and the influence weights of different types of data on the precipitation of the research area are embodied by using the ground monitoring result as a constraint, wherein the weights are important basis for precipitation calculation. And establishing a dynamic Bayesian model, and reflecting the effect of the model through ground monitoring data, namely the influence of different types of data on the estimation result, wherein the influence is reflected through the mean value and the variance of the Bayesian network model.
In step 4, the established Bayesian rainfall prediction model is as follows:
wherein p (y | D) is the posterior probability of the fused target rainfall;
y is fused target precipitation;
d is ground observation rainfall data in a certain period of time;
p(fki D) is the posterior probability of the rainfall inverted by the data under the condition of ground rainfall observation, and is also regarded as the accuracy of the rainfall inverted by different satellites under the condition of ground rainfall observation;
fkprecipitation for inversion of different models;
p(y|fkd) inverting precipitation based on different models, and observing the posterior distribution of the fusion precipitation under the precipitation condition on the ground;
k is different model types;
m is the total amount of the model;
the mean value thereof is expressed as
Wherein:
e (y | D) represents the mean of the Bayesian multi-model ensemble, wherein the E symbol represents the mean;
E[pk(y|fk,D)]the mean value of the posterior distribution of the fusion precipitation under the condition of the ground observation precipitation is used for inverting the precipitation based on different models;
wkthe shorthand of the posterior probability of rainfall is inverted for data under the condition of ground observation rainfall;
wk＝p(fki D), then have
The variance is expressed as:
wherein:
var (y | D) represents the variance based on dynamic Bayesian multi-model integration;
σ2is the variance of satellite precipitation relative to ground precipitation.
Step 5, extracting a target dynamic training sample: aiming at the Bayes rainfall prediction model established in the step 4, training the Bayes rainfall prediction model established in the step 4 by using the rainfall data observed by the ground station as a training sample to obtain a trained Bayes rainfall prediction model;
in step 6, relevant parameters (such as weight coefficients of various types of data) of the dynamic Bayesian model and a fusion result cannot be directly given, and an optimal solution needs to be obtained through continuous iteration in the implementation process. And solving the optimal solution by adopting a maximum entropy method. The corresponding conditional probability when the conditional entropy is maximum is the required conditional probability.
When the maximum entropy method is adopted to solve the nonlinear optimal solution of the trained Bayes rainfall prediction model, the corresponding conditional probability when the conditional entropy is maximum is the required conditional probability, namely:
wherein:
h (p) is the conditional entropy when the fusion rainfall and the observation correlation of the ground station are maximum;
and p (D) is the fusion of the rainfall amount as the ground observation value.
And 7, generating an estimation result of the rainfall amount fused by the multi-source information in the research area based on the optimal weight and uncertainty of the satellite data source.
And calculating the multi-source information precipitation estimation value by using the solving result obtained in the process. And obtaining the normalized weight of each type of data in the research area by interpolation according to the dynamic weight of each type of data obtained by the maximum entropy algorithm, and further comprehensively utilizing the multi-source information to calculate the precipitation of the research area.
The step 7 specifically comprises the following steps:
interpolating the optimal weight and uncertainty information of a fixed station in a target period to the whole research area by using a geostatistical spatial interpolation method, and carrying out normalization processing on the interpolated weight to ensure that the sum of the weights of each satellite of each grid point is always equal to 1, thereby obtaining the normalized weight of various satellite data in the research area and obtaining a weight spatial distribution map of the research area; and generating an estimation result of applying multi-source information fusion rainfall in the research area based on the weight space distribution map.
One embodiment is described below:
the specific embodiment discloses a rainfall estimation method based on multi-source information fusion, which comprises the following steps:
step 1: the method comprises the following steps of preprocessing multi-source satellite rainfall data:
first, a ground station in the selected area of interest observes rainfall data and suitable satellite rainfall data sources, with alternative satellite data sources including TRMM 3B42RT, TRMM 3B42V7, GPM IMERG, PERSIANN, GsMap, CPC CMORPH, and FY-2E formats.
Then, preprocessing the satellite rainfall data, decoding according to respective formats, eliminating invalid values and illegal values in the data,
and finally, checking the coordinate grid range and the effective time interval of rainfall data of each satellite, and confirming that the research area is in the control range of the selected data and the time interval meets the working requirement.
Step 2: the method comprises the following steps of constructing a multi-source satellite rainfall data volume, wherein the main work comprises the following steps of format unification, scale matching and region cutting:
the format is unified: the formats of the respective data are unified into a binary format set by the work, so that the data can be conveniently and quickly read and stored. The header file comprises a coordinate range and a data source, and the file body comprises a rainfall data matrix at a corresponding position.
And (3) scale matching: the scale matching process comprises time-of-day unification and spatial scale unification. The monitoring time and the spatial resolution of various satellite data are different, and the selected data is converted into data with uniform time and consistent spatial grid. For example, satellite data can be uniformly converted into grid data with a time interval of 2 hours and a uniform spatial position and an interval of 0.2 degrees, and grids at different moments form a multisource satellite rainfall data volume.
Area cutting: and cutting out multi-source satellite rainfall data belonging to a research area range from the multi-source satellite rainfall data, wherein the space range of the cut-out data is a rectangle, and the coordinate range of the cut-out data is determined by an extreme value of the coordinate of the research area, namely the cut-out area is a minimum rectangle containing the research area. Form a research area multi-source satellite rainfall data volume
And step 3: constructing a multi-source data set, specifically as follows:
combining the rainfall data observed by the ground station in the research area with the rainfall data of the multisource satellite in the research area to form a multisource data set in the research area.
Specifically, the rainfall data observed by the ground stations is put into corresponding grids of a multisource satellite rainfall data volume in a research area, and if a plurality of ground stations of the same ground station exist in a single grid, the average value is taken; taking grids with ground station rainfall observation data as reference points and constraint conditions, and evaluating and optimizing the conformity of the optimal solution of the Bayesian rainfall prediction model; and for grids without the data of the rainfall observed by the ground station, estimating by using an optimized Bayesian rainfall prediction model to obtain an optimal solution. The distribution and control range of the ground stations play an important role in the estimation result of the rainfall in the research area.
And 4, step 4: the model construction based on the dynamic Bayes is as follows
According to bayesian theory, when the likelihood function is 1,
that is, the bayesian model is a process in which a weight (prior probability density p (y | D)) converts satellite observation data (f) as a prior probability into ground station observation rainfall (D) as a posterior probability. Wherein the sum of the probability densities of the data of each observation station is 1
The mean value of the rainfall model based on the dynamic Bayesian principle is
The variance of this model is
The core work of the invention is to develop work aiming at the rainfall model based on the dynamic Bayesian principle to obtain the optimal weight and uncertainty information of various satellite data sources.
And 5: target dynamic training sample extraction
For the model, the rainfall data observed by a ground station is used as a training sample. Aiming at the weather characteristics of rainfall, the relation between the early rainfall and the current rainfall is utilized, for each station, 20 days before the target time and 20 days in the same time period of two years before the target time are selected, and the total number of 80 samples is used as a training sample for the station to fuse the weight. 80% of ground observation stations in the target area are selected as calibration stations to participate in multi-source information fusion rainfall estimation, and the other 20% of ground observation stations are used as independent stations to verify the estimation effect.
Step 6: entropy maximization method non-linear optimization
And (3) optimizing the weight (probability density) of each satellite information by an entropy maximization method on the basis of the rainfall information of the calibration station.
Wherein: max (H (y | D)) is the maximum value of the conditional entropy of the fusion rainfall and ground station observations;
the learning model is in the form of
Wherein Z is a normalization factor.
Pw(y | D) a maximum entropy model representing target rainfall and ground observed rainfall;
z (y) a normalization factor representing a target precipitation;
the optimization steps are as follows:
1) assuming that the probability of the rainfall data of various satellites in the initial model of the initial iteration appearing in the model is the same, namely equal probability uniform distribution;
2) estimating the weight of each satellite data in rainfall estimation by using the model of the Nth iteration, if the result exceeds the actual value detected by the ground station, reducing the weight of the corresponding data source, otherwise, increasing the weight of the corresponding data source;
3) and repeating the step 2 until the model converges.
And continuously optimizing parameters of the probability density functions of different satellite rainfalls in the training sample period to obtain the optimal weight and uncertainty information distribution of various satellite rainfalls at a certain moment of the training sample. The optimal weight and uncertainty information obtained by the scheme change along with the change of the space position and time, and the dynamic property is presented.
And 7: the spatial weight distribution is determined as follows
The main work is spatial weight and uncertainty normalization, and further spatial distribution of various satellite data weights is obtained. And interpolating the optimal weight and the uncertain information of the fixed station in the target period to the whole target space range by using a geostatistical spatial interpolation method, and carrying out normalization processing on the interpolated weight so that the sum of the weights of each satellite at each lattice point is always equal to 1.
The geostatistical interpolation method comprises a plurality of specific interpolation methods to realize two-dimensional plane local area parameter interpolation, and a specific method in the geostatistical spatial interpolation method can be selected according to the distribution condition of the ground station, such as simple kriging, common kriging, collaborative kriging or random modeling. And realizing spatial interpolation of each data volume weight, wherein the judgment standard of the interpolation result is 20% of rainfall data of the ground observation station as the independent station, and the interpolation mode with the highest conformity with the independent station is selected.
Taking simple kriging interpolation as an example, the process includes
(1) Establishing a predicted position weight data body for storing the weight of various types of rainfall data without a ground station grid;
(2) generating a variance function and a covariance function for estimating statistical correlations between the cell values, the variance function and the covariance function also depending on the autocorrelation model;
(3) predicting the value of an unknown point, and optimizing the variation function by using a global optimal method to obtain the weight of various satellite data in each target grid;
(4) evaluating the predicted deviation;
therefore, the optimal weight and uncertain information are obtained for the lattice points without station information, and the weight and uncertain information space distribution maps of different satellite products at different moments are obtained. And carrying out weighted summation on each grid point by using the weight spatial distribution map so as to obtain the rainfall fusion product in the whole spatial range.
And 8: multi-source data fusion result estimation and evaluation
And aiming at the multi-source data sets in all grids in the target area, obtaining the fusion rainfall and uncertainty at a specified moment by using the optimal weight and uncertainty information of corresponding multi-source satellite data obtained by interpolation and a weighting and summing mode. And for the grid unit with the station, taking the station rainfall result as the multi-source multi-platform fusion rainfall.
If more than 1 station is contained in the same grid, the average value of the station data is regarded as the fusion rainfall. And for the grid unit not comprising the station, carrying out multi-source multi-platform fusion rainfall estimation by adopting various data weights obtained by optimizing a Bayesian model. The multi-source fusion rainfall estimation is carried out by utilizing five satellite data (3B43RT, 3B42V7, CMORPH RAW, PERSIAN-CDR and OOR) and 102 ground station rainfall data during the Beijing area 2015.06.01-2015.08.31. The actual monitoring results of a single independent station are compared with the multi-source information estimation results as shown in fig. 2. Wherein, the solid line is an observed value of the independent station, the curve is a multi-source data estimation value, and the two have better correlation.
The evaluation results are evaluated on the basis of the individual stations. The statistical rules of the multi-source information estimation results and the estimation results of a single satellite are shown in fig. 3 and fig. 4. The DBMA term is a multi-information fusion estimation result, and both an closeness (Enclid) value and a Root Mean Square Error (RMSE) value of the DBMA term are superior to a single-satellite rainfall data result, so that the effectiveness of the method is proved.
In summary, the rainfall estimation method based on multi-source information fusion provided by the invention utilizes multi-source multi-platform rainfall monitoring information to form rainfall spatial distribution data; the rainfall characteristics are analyzed by establishing a Bayesian model suitable for multi-source data, and the rainfall in a certain period is estimated by combining historical rainfall information. Extracting the most effective rainfall data in the area by sorting and analyzing the multi-source rainfall information; and integrating and preprocessing different types of data with different precisions to obtain a multi-information data body at the same position in the region. The method is based on Bayesian theory to fuse and analyze the data volume to obtain rainfall estimation information of the area, so that the effective part of multi-source information is mined to the maximum extent and the influence of single-class data deviation is reduced. And carrying out optimization solution on the dynamic Bayesian model through a maximum entropy theory. The result of fusion analysis of multiple data sources reduces the uncertainty of regional rainfall estimation caused by the inaccuracy of a single rainfall information; the method provides more reliable data input and richer and more refined modeling data for strengthening regional high-precision disaster early warning, avoiding flood risks or estimating torrential rain and flood in a small watershed.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the principle of the present invention, and such modifications and improvements should also be considered within the scope of the present invention.
Claims (2)
1. A rainfall estimation method based on multi-source information fusion is characterized by comprising the following steps:
step 1, acquiring multi-source satellite rainfall data and ground station rainfall observation data in a research area;
step 2, preprocessing the multi-source satellite rainfall data, comprising: carrying out format unification, region cutting and scale matching on the multi-source satellite rainfall data to form a research region multi-source satellite rainfall data volume;
step 3, combining the rainfall data observed by the ground station in the research area with the rainfall data of the multisource satellite in the research area to form a multisource data set of the research area;
step 4, adopting the research area multi-source data set, utilizing the ground station observation rainfall data as a constraint, reflecting the influence weight of satellite rainfall data from different sources on the rainfall of the research area, and further establishing a Bayesian rainfall prediction model of the research area based on a dynamic Bayesian theory; the Bayesian rainfall prediction model is a process of converting multisource satellite rainfall data serving as prior probability into ground station observation rainfall data serving as posterior probability by weight;
step 5, extracting a target dynamic training sample: aiming at the Bayes rainfall prediction model established in the step 4, training the Bayes rainfall prediction model established in the step 4 by using the rainfall data observed by the ground station as a training sample to obtain a trained Bayes rainfall prediction model;
step 6, solving the nonlinear optimal solution of the trained Bayes rainfall prediction model obtained in the step 5 by using a maximum entropy method, and further determining the optimal weight and uncertainty information of each satellite data source; the optimal weight and the uncertainty change along with the change of the space position and the time, and the dynamic property is presented;
step 7, generating an estimation result of the rainfall amount of the applied multi-source information fusion in the research area based on the optimal weight and uncertainty of the satellite data source;
in step 2, the specific process of format unification is as follows: unifying the formats of the multi-source satellite rainfall data into a binary format; the header file comprises a coordinate range and a data source, and the file body comprises a rainfall data matrix at a corresponding position;
the specific process of the region cutting is as follows: cutting out multi-source satellite rainfall data belonging to a research area range from the multi-source satellite rainfall data, wherein the space range of the cut-out data is a rectangle, and the coordinate range of the cut-out data is determined by an extreme value of a research area coordinate, namely the cut-out area is a minimum rectangle containing the research area;
the specific process of the scale matching is as follows: the scale matching process comprises time unification and space unification; because the monitoring time and the spatial resolution capability of various satellite rainfall data are different, the multisource satellite rainfall data in the research area range are converted into data with uniform time and uniform spatial grids;
converting multi-source satellite rainfall data into a data body with uniform format, consistent time interval and space position through format unification, region cutting and scale matching, thereby forming a multi-source satellite rainfall data body of rainfall at each moment in a standard grid in a research region;
wherein, the step 3 is specifically as follows:
putting the rainfall data observed by the ground stations into corresponding grids of a multisource satellite rainfall data volume in a research area, and taking an average value if a plurality of ground stations of the same ground station exist in a single grid for observing the rainfall data; taking grids with ground station rainfall observation data as reference points and constraint conditions, and evaluating and optimizing the conformity of the optimal solution of the Bayesian rainfall prediction model; for grids without ground station rainfall observation data, estimating by using an optimized Bayesian rainfall prediction model to obtain an optimal solution;
in step 4, the established bayesian rainfall prediction model is as follows:
wherein p (y | D) is the posterior probability of the fused target rainfall;
y is fused target precipitation;
d is ground observation rainfall data in a certain period of time;
p(fki D) is the posterior probability of the rainfall inverted by the data under the condition of ground rainfall observation, and is also regarded as the accuracy of the rainfall inverted by different satellites under the condition of ground rainfall observation;
fkprecipitation for inversion of different models;
p(y|fkd) inverting precipitation based on different models, and observing the posterior distribution of the fusion precipitation under the precipitation condition on the ground;
k is different model types;
m is the total amount of the model;
the mean value thereof is expressed as
Wherein:
e (y | D) represents the mean of the Bayesian multi-model ensemble, wherein the E symbol represents the mean;
E[pk(y|fk,D)]the mean value of the posterior distribution of the fusion precipitation under the condition of the ground observation precipitation is used for inverting the precipitation based on different models;
wkthe shorthand of the posterior probability of rainfall is inverted for data under the condition of ground observation rainfall;
wk＝p(fki D), then have
The variance is expressed as:
wherein:
var (y | D) represents the variance based on dynamic Bayesian multi-model integration;
σ2is the variance of the satellite precipitation relative to the ground precipitation;
in step 6, when the maximum entropy method is adopted to solve the nonlinear optimal solution of the trained bayesian rainfall prediction model, the corresponding conditional probability when the conditional entropy is maximum is the conditional probability which needs to be obtained, that is:
wherein:
h (p) is the conditional entropy when the fusion rainfall and the observation correlation of the ground station are maximum;
p (D) fusing the rainfall into a ground observation value;
wherein, the step 7 specifically comprises the following steps: interpolating the optimal weight and uncertainty information of a fixed station in a target period to the whole research area by using a geostatistical spatial interpolation method, and carrying out normalization processing on the interpolated weight to ensure that the sum of the weights of each satellite of each grid point is always equal to 1, thereby obtaining the normalized weight of various satellite data in the research area and obtaining a weight spatial distribution map of the research area; based on the weight spatial distribution map, generating an estimation result of applying multi-source information fusion rainfall in the research area;
the method comprises the following specific steps:
step 1: the method comprises the following steps of preprocessing multi-source satellite rainfall data:
firstly, selecting ground station observation rainfall data and a proper satellite rainfall data source in a research area, wherein the selected satellite data source comprises TRM 3B42RT, TRMM 3B42V7, GPM I MERG, PERSI ANN, Gs Map, CPC CMORPH and FY-2E formats;
then, preprocessing the satellite rainfall data, decoding according to respective formats, eliminating invalid values and illegal values in the data,
finally, checking the coordinate grid range and the effective time interval of rainfall data of each satellite, and confirming that the research area is in the control range of the selected data and the time interval meets the working requirement;
step 2: the method comprises the following steps of constructing a multi-source satellite rainfall data volume, wherein the main work comprises the following steps of format unification, scale matching and region cutting:
the format is unified: the formats of the respective data are unified into a binary format set by the work, so that the data can be conveniently and quickly read and stored; the header file comprises a coordinate range and a data source, and the file body comprises a rainfall data matrix at a corresponding position;
and (3) scale matching: the scale matching process comprises time unification and space unification; the monitoring time and the spatial resolution of various satellite data are different, and the selected data is converted into data with uniform time and consistent spatial grids; the satellite data can be uniformly converted into grid data with time interval of 2 hours, uniform spatial positions and 0.2-degree interval, and grids at different moments form a multi-source satellite rainfall data body;
area cutting: cutting out multi-source satellite rainfall data belonging to a research area range from the multi-source satellite rainfall data, wherein the space range of the cut-out data is a rectangle, and the coordinate range of the cut-out data is determined by an extreme value of a research area coordinate, namely the cut-out area is a minimum rectangle containing the research area; form a research area multi-source satellite rainfall data volume
And step 3: constructing a multi-source data set, specifically as follows:
combining rainfall data observed by a ground station in a research area with rainfall data of multisource satellites in the research area to form a multisource data set of the research area;
specifically, the rainfall data observed by the ground stations is put into corresponding grids of a multisource satellite rainfall data volume in a research area, and if a plurality of ground stations of the same ground station exist in a single grid, the average value is taken; taking grids with ground station rainfall observation data as reference points and constraint conditions, and evaluating and optimizing the conformity of the optimal solution of the Bayesian rainfall prediction model; for grids without ground station rainfall observation data, estimating by using an optimized Bayesian rainfall prediction model to obtain an optimal solution; the distribution and control range of the ground stations play an important role in the estimation result of the rainfall in the research area;
and 4, step 4: the model construction based on the dynamic Bayes is as follows
According to bayesian theory, when the likelihood function is 1,
the Bayesian model is a weight, namely a prior probability density p (y | D), and the process of converting the satellite observation data serving as the prior probability into the ground station observation rainfall serving as the posterior probability; wherein the sum of the probability densities of the data of each observation station is 1
The mean value of the rainfall model based on the dynamic Bayesian principle is
The variance of this model is
Working is carried out aiming at the rainfall model based on the dynamic Bayesian principle to obtain the optimal weight and uncertainty information of various satellite data sources;
and 5: target dynamic training sample extraction
Aiming at the model, observing rainfall data by using a ground station as a training sample; aiming at the weather characteristics of rainfall, the relation between the early rainfall and the current rainfall is utilized, for each station, 20 days before the target time and 20 days in the same time period of two years before the target time are selected, and the total number of 80 samples is used as a training sample for the station fusion weight; 80% of ground observation stations in the target area are selected as calibration stations to participate in multi-source information fusion rainfall estimation, and the other 20% of ground observation stations are used as independent stations to verify the estimation effect;
step 6: entropy maximization method non-linear optimization
Taking rainfall information of a calibration station as a basis, and optimizing the weight, namely probability density, of each satellite information by an entropy maximization method;
wherein: max (H (y | D)) is the maximum value of the conditional entropy of the fusion rainfall and ground station observations;
the learning model is in the form of
Wherein z is a normalization factor;
Pw(y | D) a maximum entropy model representing target rainfall and ground observed rainfall;
z (y) a normalization factor representing a target precipitation;
the optimization steps are as follows:
1) assuming that the probability of the rainfall data of various satellites in the initial model of the initial iteration appearing in the model is the same, namely equal probability uniform distribution;
2) estimating the weight of each satellite data in rainfall estimation by using the model of the Nth iteration, if the result exceeds the actual value detected by the ground station, reducing the weight of the corresponding data source, otherwise, increasing the weight of the corresponding data source;
3) repeating the step 2 until the model is converged;
continuously optimizing parameters of probability density functions of rainfall of different satellites in a training sample period to obtain optimal weight and distribution of uncertainty information of rainfall of various satellites at a certain moment of the training sample; the optimal weight and uncertainty information obtained by the scheme change along with the change of the space position and time, and the dynamic property is presented;
and 7: the spatial weight distribution is determined as follows
The main work is space weight and uncertainty normalization, so that the space distribution of various satellite data weights is obtained; interpolating the optimal weight and uncertain information of a fixed station in a target period to the whole target space range by using a geostatistical spatial interpolation method, and carrying out normalization processing on the interpolated weight to ensure that the sum of the weights of each satellite of each grid point is always equal to 1;
the geological statistical interpolation method comprises a plurality of specific interpolation methods to realize two-dimensional plane local area parameter interpolation, and specific methods in the geological statistical spatial interpolation method can be selected according to the distribution condition of the ground station, wherein the specific methods comprise a simple kriging, a common kriging, a collaborative kriging or a random modeling mode; spatial interpolation of each data volume weight is realized, the judgment standard of an interpolation result is 20% of rainfall data of a ground observation station as an independent station, and an interpolation mode with the highest conformity with the independent station is selected;
for simple kriging interpolation, the process includes
(1) Establishing a predicted position weight data body for storing the weight of various types of rainfall data without a ground station grid;
(2) generating a variance function and a covariance function for estimating statistical correlations between the cell values, the variance function and the covariance function also depending on the autocorrelation model;
(3) predicting the value of an unknown point, and optimizing the variation function by using a global optimal method to obtain the weight of various satellite data in each target grid;
(4) evaluating the predicted deviation;
therefore, the optimal weight and uncertain information are obtained for the lattice points without station information, and the weight and uncertain information space distribution maps of different satellite products at different moments are obtained; weighting and summing each grid point by using the weight space distribution map so as to obtain rainfall fusion products in the whole space range;
and 8: multi-source data fusion result estimation and evaluation
Aiming at multi-source data sets in all grids in a target area, obtaining the fusion rainfall and uncertainty at a specified time by using the optimal weight and uncertainty information of corresponding multi-source satellite data obtained by interpolation and in a weighted summation mode; for grid units with stations, station rainfall results are used as multi-source multi-platform fusion rainfall;
if more than 1 station is included in the same grid, the average value of the station data is regarded as the fusion rainfall; and for the grid unit not comprising the station, carrying out multi-source multi-platform fusion rainfall estimation by adopting various data weights obtained by optimizing a Bayesian model.
2. The rainfall estimation method based on multi-source information fusion of claim 1, wherein the multi-source satellite rainfall data is rainfall data from different platforms and different satellite types; after original multi-source satellite rainfall data are obtained, abnormal values are removed, and then format unification, area cutting and scale matching processing are carried out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810427354.2A CN108761574B (en) | 2018-05-07 | 2018-05-07 | Rainfall estimation method based on multi-source information fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810427354.2A CN108761574B (en) | 2018-05-07 | 2018-05-07 | Rainfall estimation method based on multi-source information fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108761574A CN108761574A (en) | 2018-11-06 |
| CN108761574B true CN108761574B (en) | 2021-04-27 |
Family
ID=64010034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810427354.2A Active CN108761574B (en) | 2018-05-07 | 2018-05-07 | Rainfall estimation method based on multi-source information fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108761574B (en) |
Families Citing this family (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109697323B (en) * | 2018-12-27 | 2022-08-16 | 武汉大学 | Rainfall observation method integrating satellite remote sensing and mobile communication base station signals |
| CN109884735B (en) * | 2019-03-27 | 2019-11-19 | 山东省气象局大气探测技术保障中心 | A method of precipitation is observed based on rainy quantity sensor |
| CN110118982B (en) * | 2019-04-12 | 2022-11-25 | 大连理工大学 | Satellite precipitation data correction method based on space optimization interpolation |
| CN110457374B (en) * | 2019-08-20 | 2023-04-18 | 贵州东方世纪科技股份有限公司 | Method for identifying typical rainstorm process in time period |
| CN110502598A (en) * | 2019-08-22 | 2019-11-26 | 华东师范大学 | Intelligent independent positioning and space-time geodata fusion method |
| CN112766531B (en) * | 2019-11-06 | 2023-10-31 | 中国科学院国家空间科学中心 | Runoff prediction system and method based on satellite microwave observation data |
| CN110991702B (en) * | 2019-11-13 | 2021-02-12 | 清华大学 | Method and device for calculating rainfall in mountainous area, computer equipment and storage medium |
| CN110674127A (en) * | 2019-11-14 | 2020-01-10 | 湖南国天电子科技有限公司 | Ocean sediment test system and method based on deep learning |
| CN110865425B (en) * | 2019-11-28 | 2021-10-26 | 中国有色金属长沙勘察设计研究院有限公司 | Rain gauge gross error detection method based on prior information |
| CN111126466B (en) * | 2019-12-16 | 2020-09-22 | 西安科技大学 | Multi-source PWV data fusion method |
| CN111199345B (en) * | 2019-12-27 | 2023-07-11 | 河北建筑工程学院 | Method for measuring and calculating design rainfall of sponge city and terminal equipment |
| CN111582546B (en) * | 2020-04-09 | 2023-07-25 | 中国科学院国家空间科学中心 | Method and system for acquiring global rainfall information in different places by utilizing rainfall data set |
| CN111638565A (en) * | 2020-06-09 | 2020-09-08 | 云南电网有限责任公司电力科学研究院 | Method for monitoring rainstorm in mountainous area |
| CN113837422A (en) * | 2020-06-24 | 2021-12-24 | 海南省气象科学研究所 | Precipitation falling area determination method based on multi-source big data |
| CN111624682B (en) * | 2020-06-24 | 2021-11-30 | 海南省气象科学研究所 | Quantitative precipitation estimation method based on multi-source data fusion |
| CN111832828B (en) * | 2020-07-17 | 2023-12-19 | 国家卫星气象中心(国家空间天气监测预警中心) | Intelligent precipitation prediction method based on wind cloud No. four meteorological satellites |
| CN111983732B (en) * | 2020-07-27 | 2022-08-05 | 南京信息工程大学 | Rainfall intensity estimation method based on deep learning |
| CN111859303B (en) * | 2020-07-28 | 2023-09-05 | 南京大学 | Soil humidity fusion method and system based on dynamic Bayesian average |
| CN112099016A (en) * | 2020-11-09 | 2020-12-18 | 中国水利水电科学研究院 | Heaven-earth collaborative flood situation emergency monitoring system and method |
| CN112308029A (en) * | 2020-11-24 | 2021-02-02 | 国网湖南省电力有限公司 | Rainfall station and satellite rainfall data fusion method and system |
| CN112685492B (en) * | 2020-12-30 | 2024-04-16 | 北京墨迹风云科技股份有限公司 | Multi-source data fusion method and device |
| CN112381337B (en) * | 2021-01-14 | 2021-04-23 | 国能大渡河大数据服务有限公司 | Multi-source meteorological data fusion processing method, system, terminal and medium |
| CN112800634B (en) * | 2021-04-07 | 2021-06-25 | 水利部交通运输部国家能源局南京水利科学研究院 | Rainfall estimation method and system coupling dry-wet state identification and multi-source information fusion |
| CN113205155A (en) * | 2021-05-27 | 2021-08-03 | 中国水利水电科学研究院 | Multi-source precipitation data fusion method based on partition self-adaptive weight |
| CN113408019B (en) * | 2021-06-23 | 2022-04-22 | 河北地质大学 | Water system sediment geochemical anomaly mapping method based on BME-GWR |
| CN113673777B (en) * | 2021-08-27 | 2024-02-02 | 华北电力大学 | Desert succession prediction method under climate change condition |
| CN114970743B (en) * | 2022-06-17 | 2022-11-08 | 中国科学院地理科学与资源研究所 | Multi-source remote sensing rainfall data fusion method based on multi-modal deep learning |
| CN114791638B (en) * | 2022-06-29 | 2022-09-16 | 山东大学 | Super-heavy rainstorm monitoring method and system based on multi-source data |
| CN116151483B (en) * | 2023-04-14 | 2023-07-21 | 山东理工大学 | Regional rainfall landslide probabilistic prediction method and prediction terminal |
| CN117033935B (en) * | 2023-08-03 | 2024-02-23 | 北京市市政工程设计研究总院有限公司 | Prediction method of rainfall characteristic under statistics and monitoring based on Bayesian fusion |
| CN117081161B (en) * | 2023-10-17 | 2024-01-26 | 长江三峡集团实业发展(北京)有限公司 | Method, device, computer equipment and medium for determining water-light complementary optimization target |
| CN117150437B (en) * | 2023-10-31 | 2024-01-30 | 航天宏图信息技术股份有限公司 | Multi-source satellite sea surface wind field data fusion method, device, equipment and medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1847856A1 (en) * | 2006-04-21 | 2007-10-24 | Fondazione Torino Wireless | Method and system for forecasting an ambient variable |
| CN103472501A (en) * | 2013-09-06 | 2013-12-25 | 中国气象科学研究院 | Cloud and aerial total cloud amount detection method and system |
| CN106991278A (en) * | 2017-03-21 | 2017-07-28 | 武汉大学 | It is a kind of to gather precipitation forecast and the coupling process of real-time flood probability forecast |
| CN107918166A (en) * | 2016-10-09 | 2018-04-17 | 清华大学 | More satellite fusion precipitation methods and system |
| CN107918165A (en) * | 2016-10-09 | 2018-04-17 | 清华大学 | More satellites fusion Prediction of Precipitation method and system based on space interpolation |
-
2018
- 2018-05-07 CN CN201810427354.2A patent/CN108761574B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1847856A1 (en) * | 2006-04-21 | 2007-10-24 | Fondazione Torino Wireless | Method and system for forecasting an ambient variable |
| CN103472501A (en) * | 2013-09-06 | 2013-12-25 | 中国气象科学研究院 | Cloud and aerial total cloud amount detection method and system |
| CN107918166A (en) * | 2016-10-09 | 2018-04-17 | 清华大学 | More satellite fusion precipitation methods and system |
| CN107918165A (en) * | 2016-10-09 | 2018-04-17 | 清华大学 | More satellites fusion Prediction of Precipitation method and system based on space interpolation |
| CN106991278A (en) * | 2017-03-21 | 2017-07-28 | 武汉大学 | It is a kind of to gather precipitation forecast and the coupling process of real-time flood probability forecast |
Non-Patent Citations (2)
| Title |
|---|
| 基于信息熵的降雨信息区域化分析;张继国，吴敏，谢平等;《河海大学学报（自然科学版）》;20131130;第41卷(第6期);第477，478页 * |
| 基于最大熵谱和云模型的巢湖流域降水预测;吕军，程先富，孙鸿鹄;《自然灾害学报》;20170228;第26卷(第1期);第51,52页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108761574A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108761574B (en) | Rainfall estimation method based on multi-source information fusion | |
| US10852439B1 (en) | Global ionospheric total electron content prediction system | |
| US20220043182A1 (en) | Spatial autocorrelation machine learning-based downscaling method and system of satellite precipitation data | |
| Schuurmans et al. | Automatic prediction of high-resolution daily rainfall fields for multiple extents: The potential of operational radar | |
| CN112800634B (en) | Rainfall estimation method and system coupling dry-wet state identification and multi-source information fusion | |
| CN109946762B (en) | Method and system for short-time rainfall forecast based on probability distribution | |
| Sartini et al. | Comparing different extreme wave analysis models for wave climate assessment along the Italian coast | |
| Arowolo et al. | Comparison of spatial interpolation techniques to generate high‐resolution climate surfaces for Nigeria | |
| CN113496104A (en) | Rainfall forecast correction method and system based on deep learning | |
| CN113901384A (en) | Ground PM2.5 concentration modeling method considering global spatial autocorrelation and local heterogeneity | |
| CN111915158A (en) | Rainstorm disaster weather risk assessment method, device and equipment based on Flood Area model | |
| Fan et al. | A comparative study of four merging approaches for regional precipitation estimation | |
| CN115357847A (en) | Day scale star-ground precipitation fusion method based on error decomposition | |
| Siqueira et al. | Postprocessing continental-scale, medium-range ensemble streamflow forecasts in South America using Ensemble Model Output Statistics and Ensemble Copula Coupling | |
| CN114936201A (en) | Satellite precipitation data correction method based on adaptive block neural network model | |
| Cheng et al. | Geostatistical interpolation of space–time rainfall on Tamshui River basin, Taiwan | |
| Xiang et al. | Real-Time streamflow forecasting framework, implementation and post-analysis using deep learning | |
| CN115829163B (en) | Multi-mode integration-based runoff prediction method and system for middle and lower reaches of Yangtze river | |
| Tang et al. | Merging daily sea surface temperature data from multiple satellites using a Bayesian maximum entropy method | |
| Hema et al. | Reconstructing missing hourly real-time precipitation data using a novel intermittent sliding window period technique for automatic weather station data | |
| CN117171128A (en) | Aquatic organism protection threshold identification method based on four-water coupling model | |
| Liu et al. | Gridded statistical downscaling based on interpolation of parameters and predictor locations for summer daily precipitation in North China | |
| Capsoni et al. | 1-min rain rate statistics predictions from 1-hour rain rate statistics measurements | |
| Coccia et al. | Recent developments in predictive uncertainty assessment based on the model conditional processor approach. | |
| Neykov et al. | Linking atmospheric circulation to daily precipitation patterns over the territory of Bulgaria |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |