CN107918166B - More satellite fusion precipitation methods and system - Google Patents
More satellite fusion precipitation methods and system Download PDFInfo
- Publication number
- CN107918166B CN107918166B CN201610881113.6A CN201610881113A CN107918166B CN 107918166 B CN107918166 B CN 107918166B CN 201610881113 A CN201610881113 A CN 201610881113A CN 107918166 B CN107918166 B CN 107918166B
- Authority
- CN
- China
- Prior art keywords
- precipitation
- satellite
- data
- precipitation data
- fusion
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/14—Rainfall or precipitation gauges
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Hydrology & Water Resources (AREA)
- Engineering & Computer Science (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The present invention provides a kind of more satellites fusion precipitation method and system, the method comprise the steps that obtaining sampling periods, the sampling periods include each year in N preset period of time, and the duration of each preset period of time was less than 1 year；The precipitation data of the surface precipitation data and at least two satellites in the sampling periods is obtained respectively；Determine the prior probability of each satellite；According to the prior probability of each satellite, the precipitation data of the surface precipitation data and each satellite, more satellites are calculated using probability statistics model and merge precipitation parameter.More satellites provided by the present invention merge precipitation method, according to the difference of sampling periods, and the difference of the ground station selection, the calculated result of the fusion weight coefficient of each satellite is not also identical, every satellite error as caused by monitoring time or the difference in space is excluded to the greatest extent, so that fused precipitation result is more accurate.
Description
Technical field
The present invention relates to satellites to merge precipitation technical field, merges precipitation method and system more particularly to more satellites.
Background technique
Precipitation is one of most critical input parameter of atmosphere hydrology model uncertainty, and global moisture and energy circulation
Core component.In traditional technology, obtaining precipitation data, there are mainly three types of modes.First is that ground station is observed, using office
The result of portion's point sampling observation represents the true precipitation within the scope of even several hundred square kilometres of periphery tens.It is now recognized that earth station
The precipitation data of point is the precipitation measurement data trusted the most, but the observation data of ground station exist obviously in spatial and temporal distributions
Discontinuity, it is difficult to the significant Spatial-Temporal Variability of the practical precipitation of reflection.On the other hand, Chinese ground rainfall gauge website distribution east
Mi Xi dredge, observational data spatio-temporal distribution is also unequal, cause based on ground station observe precipitation data quality by
Influence, therefore, although ground station observation can precise measurement surface precipitation, influenced by network density and its spatial distribution
Larger, especially for the relative complex mountain area of landform, existing station net layout is not able to satisfy application demand.Second is that weather radar,
Weather radar obtains the continuous precipitation information in space, in certain journey by physical quantity related with precipitation in atmospheric sounding indirectly
The deficiency of ground station spatial distribution is compensated on degree, but weather radar is easy to be multifactor by electronic signal and running environment etc.
It influences, such as the uncertainty of terrain shading, radar ray lifting and Z-R (radar reflectivity Z and rainfall intensity R) relationship,
Regions with complex terrain has biggish uncertainty, and it is poor that precision is observed compared with the rainfall gauge of ground station.Third is that meteorological satellite, satellite
The retrieving precipitation data of monitoring are wide with observation scope, time interval is short, for independent, discrete ground station observation
Spatial and temporal distributions are more continuous, have been increasingly becoming the weight that rainfall monitoring and hazard forecasting early warning are carried out on global and local scale
Tool is wanted, while the hydrologic research for also lacking survey data basin for ground provides the rainfall observation information of great application value.But
As weather radar, satellite remote sensing technology is also the indirect operation means to rainfall, is calculated by remote sensing instrument, inverting
The factors such as method influence, and the precision of product is relatively low, have significant uncertain.The retrieving precipitation data estimation of each satellite misses
Difference is related with the factors such as survey region, type of precipitation, rainy season, surface cover situation, landform, has in different space-time uniques
There are respective advantage and disadvantage.In order to more truly describe actual changes and precipitation, space division when needing to merge preferably capture precipitation field
The satellite Retrieval precipitation data of cloth comprehensively considers the error characteristics and observation advantage of different satellites.
More satellites fusion precipitation method more traditional at present has: simple arithmetic mean, removal maximum deviation etc..Wherein,
Simple arithmetic mean is easily understood, and is one kind quickly by the method for the retrieving precipitation data fusion of each satellite, but this method is recognized
For each Satellite Product data weight having the same at different times, the ability for capturing precipitation is identical, this and actual conditions
It is not consistent.Maximum deviation method is removed first to reject the maximum satellite Retrieval precipitation data of deviation, it is then anti-to remaining satellite
It drills precipitation data and carries out arithmetic average, consider differing for the performance of different satellite Retrieval precipitation to a certain extent, but to surplus
When remaining satellite Retrieval precipitation data carries out arithmetic average, the problem of equally existing arithmetic average method.
How according to the retrieving precipitation data of multi-satellite, more satellites fusion drop of high-precision and high-spatial and temporal resolution is obtained
Water is as a result, be satellite fusion precipitation field technical problem urgently to be resolved.
Summary of the invention
Based on this, it is necessary to for the problem that in more satellites fusion precipitation, different satellites merge weight, provide one kind and defend more
Star merges precipitation method and system.
The present invention provides a kind of more satellite fusion precipitation methods, which comprises
Sampling periods are obtained, the sampling periods include each year in N preset period of time, each preset period of time
Duration was less than 1 year, and wherein N is positive integer；
The precipitation data of the surface precipitation data and at least two satellites in the sampling periods is obtained respectively；
Determine the prior probability of each satellite；
According to the prior probability of each satellite, the precipitation data of the surface precipitation data and each satellite, utilize
Probability statistics model calculates more satellites and merges precipitation parameter, and more satellite fusion precipitation parameters include the fusion of each satellite
Weight coefficient.
The prior probability of the determination each satellite in one of the embodiments, comprising:
The prior probability of each satellite is calculated according to the precipitation data of the surface precipitation data and each satellite, or
Obtain the preset prior probability of each satellite.
It is described according to the prior probability of each satellite, the surface precipitation data and institute in one of the embodiments,
The precipitation data for stating each satellite calculates more satellites using probability statistics model and merges precipitation parameter, comprising:
According to the Posterior distrbutionp of the posterior probability of the precipitation data of each satellite and more satellites fusion precipitation, determine defend more
The probability density function of star fusion precipitation, wherein the posterior probability of the precipitation data of each satellite is the precipitation number of each satellite
According to the posterior probability under the surface precipitation data qualification, the Posterior distrbutionp of more satellite fusion precipitation is based on described each
The Posterior distrbutionp of more satellites fusion precipitation under the precipitation data of satellite and the surface precipitation data qualification；
The probability density function of more satellite fusion precipitation is iterated calculating, determines more satellite fusion precipitation
Parameter.
The probability density function by more satellite fusion precipitation is iterated meter in one of the embodiments,
It calculates, determines more satellite fusion precipitation parameters, comprising:
The precipitation data of the surface precipitation data and each satellite is subjected to normal distribution conversion respectively, obtains ground
The normal distribution precipitation data of normal distribution precipitation data and each satellite；
According to the ground normal distribution precipitation data, the normal distribution precipitation data of each satellite, defended described more
The probability density function of star fusion precipitation is iterated calculating using EM algorithm, the precipitation of each satellite after determining optimization
The posterior probability of data；
The posterior probability of the precipitation data of each satellite after the optimization is determined as to the fusion weight system of each satellite
Number.
The normal state according to the ground normal distribution precipitation data, each satellite in one of the embodiments,
It is distributed precipitation data, the probability density function of more satellite fusion precipitation is iterated calculating using EM algorithm,
The posterior probability of the precipitation data of each satellite after determining optimization, comprising:
According to the precipitation data of each satellite and the surface precipitation data, the precipitation data of each satellite is calculated
Error；
By the error of the precipitation data of each satellite, the posterior probability of the precipitation data of each satellite, it is determined as institute
State the initial parameter sets to be asked in the probability density function of more satellite fusion precipitation；
According to the set of the parameter initially to be asked, the normal state of the ground normal distribution precipitation data and each satellite
It is distributed precipitation data, determines the log-likelihood function of the parameter sets initially to be asked；
Iterate to calculate the log-likelihood function and determine the maximum likelihood value of the log-likelihood function, according to it is described most
Maximum-likelihood value determines the parameter sets to be asked after optimization, and after the optimization includes each satellite after optimizing wait seek parameter sets
The posterior probability of precipitation data.
The iterative calculation log-likelihood function and the determining log-likelihood function in one of the embodiments,
Maximum likelihood value, specifically include:
According to each preset initial weight value of satellite, the precipitation number of the surface precipitation data and each satellite
According to the initial variance of the precipitation data of calculating each satellite；
According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；
According to the precipitation data of the satellite, the variance and the surface precipitation data, hidden variable is determined；
The hidden variable is iterated calculating, the hidden variable after obtaining iteration；
According to the hidden variable after the iteration, the iteration weighted value and iteration variance of each satellite are calculated；
According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；
According to the likelihood value and preset threshold value, the maximum likelihood value of the log-likelihood function is determined.
More satellites fusion precipitation parameter of the known grid further includes more satellite fusion drops in one of the embodiments,
The calculation method of the control information of water, the control information includes:
According to the parameter sets to be asked after the optimization, the side of the probability density function of more satellite fusion precipitation is calculated
Difference；
The variance of the probability density function of more satellite fusion precipitation is determined as to the mistake of more satellite fusion precipitation
Poor information.
More satellites provided by the present invention merge precipitation method, are clicked according to the difference of sampling periods and the earth station
The calculated result of the difference selected, the fusion weight coefficient of each satellite is not also identical, the fusion weight coefficient of each satellite when
Between and dynamic characteristic is spatially presented, exclude every satellite to the greatest extent since the difference in monitoring time or space is produced
Raw error, so that fused precipitation result is more accurate.
In one embodiment of more satellite fusion precipitation methods provided by the present invention, by calculating more satellite fusion drops
The control information of water provides the error of more satellites fusion precipitation, for further correct more satellites fusion precipitation as a result, improving
The accuracy rate of more satellite fusion precipitation.
The present invention also provides a kind of more satellites to merge precipitation system, comprising:
Period sampling module, for obtaining sampling periods, the sampling periods include each year in N preset period of time,
The duration of each preset period of time was less than 1 year, and wherein N is positive integer；
Precipitation data obtains module, defends for obtaining surface precipitation data in the sampling periods and at least two respectively
The precipitation data of star；
Prior probability determining module, for determining the prior probability of each satellite；
Precipitation parameter computing module, for according to the prior probability of each satellite, surface precipitation data and described
The precipitation data of each satellite calculates more satellites using probability statistics model and merges precipitation parameter, more satellite fusion precipitation ginsengs
Number includes the fusion weight coefficient of each satellite.
The prior probability determining module in one of the embodiments, for according to the surface precipitation data and institute
The precipitation data for stating each satellite calculates the prior probability of each satellite, or obtains each preset prior probability of satellite.
The precipitation parameter computational submodule in one of the embodiments, comprising:
Probability density function determination unit, for being melted according to the posterior probability and more satellites of the precipitation data of each satellite
The Posterior distrbutionp for closing precipitation determines the probability density function of more satellite fusion precipitation, wherein the precipitation data of each satellite
Posterior probability is posterior probability of the precipitation data of each satellite under the surface precipitation data qualification, more satellite fusion drops
The Posterior distrbutionp of water is that more satellites under the precipitation data based on each satellite and the surface precipitation data qualification merge drop
The Posterior distrbutionp of water；
Precipitation parameter computing unit is merged, based on being iterated the probability density function of more satellite fusion precipitation
It calculates, determines more satellite fusion precipitation parameters.
The fusion precipitation parameter computing unit in one of the embodiments, comprising:
Normal distribution conversion subunit, for by the precipitation data of the surface precipitation data and each satellite respectively into
Row normal distribution conversion, obtains the normal distribution precipitation data of ground normal distribution precipitation data and each satellite；
Greatest hope computation subunit, for the normal state according to the ground normal distribution precipitation data, each satellite
It is distributed precipitation data, the probability density function of more satellite fusion precipitation is iterated calculating using EM algorithm,
The posterior probability of the precipitation data of each satellite after determining optimization；And by the posteriority of the precipitation data of each satellite after the optimization
Determine the probability is the fusion weight coefficient of each satellite.
The greatest hope computation subunit in one of the embodiments, comprising:
Error computation component, for according to each satellite precipitation data and the surface precipitation data, described in calculating
The error of the precipitation data of each satellite；
Initially parameter obtaining component to be asked, for by the drop of the error of the precipitation data of each satellite, each satellite
The posterior probability of water number evidence, the initial parameter sets to be asked being determined as in the probability density function of more satellite fusion precipitation；
Likelihood function securing component, for the set according to the parameter initially to be asked, the ground normal distribution precipitation
The normal distribution precipitation data of data and each satellite determines the log-likelihood function of the parameter sets initially to be asked；
Maximum likelihood value securing component, for iterating to calculate the log-likelihood function and determining the log-likelihood function
Maximum likelihood value, the parameter sets to be asked after optimization, parameter to be asked after the optimization are determined according to the maximum likelihood value
Set includes the posterior probability of the precipitation data of each satellite after optimization.
The maximum likelihood value securing component in one of the embodiments, is specifically used for:
According to each preset initial weight value of satellite, the precipitation number of the surface precipitation data and each satellite
According to the initial variance of the precipitation data of calculating each satellite；
According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；
According to the precipitation data of the satellite, the variance and the surface precipitation data, hidden variable is determined；
The hidden variable is iterated calculating, the hidden variable after obtaining iteration；
According to the hidden variable after the iteration, the iteration weighted value and iteration variance of each satellite are calculated；
According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；
According to the likelihood value and preset threshold value, the maximum likelihood value of the log-likelihood function is determined.
The known grid computing module in one of the embodiments, is also used to calculate known grid more and defends
Star merges precipitation parameter, and more satellites fusion precipitation parameter of the known grid further includes the error letter of more satellite fusion precipitation
Breath；
The greatest hope computation subunit, further includes:
Variance computation module, for calculating more satellites and merging precipitation according to the parameter sets to be asked after the optimization
Probability density function variance；
Control information determines component, for the variance of the probability density function of more satellite fusion precipitation to be determined as institute
State the control information of more satellite fusion precipitation.
More satellites provided by the present invention merge precipitation system, are clicked according to the difference of sampling periods and the earth station
The calculated result of the difference selected, the fusion weight coefficient of each satellite is not also identical, the fusion weight coefficient of each satellite when
Between and dynamic characteristic is spatially presented, exclude every satellite to the greatest extent since the difference in monitoring time or space is produced
Raw error, so that fused precipitation result is more accurate.
In one embodiment of more satellite fusion precipitation systems provided by the present invention, by calculating more satellite fusion drops
The control information of water provides the error of more satellites fusion precipitation, for further correct more satellites fusion precipitation as a result, improving
The accuracy rate of more satellite fusion precipitation.
Detailed description of the invention
Fig. 1 is the flow diagram of more satellites fusion precipitation method in one embodiment；
Fig. 2 is the flow diagram of more satellites fusion precipitation method in another embodiment；
Fig. 3 is the flow diagram of more satellites fusion precipitation method in another embodiment；
Fig. 4 is the flow diagram of more satellites fusion precipitation method in further embodiment；
Fig. 5 is the flow diagram of more satellites fusion precipitation method in further embodiment；
Fig. 6 is the flow diagram of more satellites fusion precipitation method in further embodiment；
Fig. 7 is the structural schematic diagram of more satellites fusion precipitation system in one embodiment；
Fig. 8 is the structural schematic diagram of more satellites fusion precipitation system in another embodiment；
Fig. 9 is the structural schematic diagram of more satellites fusion precipitation system in another embodiment；
Figure 10 is the structural schematic diagram of more satellites fusion precipitation system in further embodiment；
Figure 11 is the structural schematic diagram of more satellites fusion precipitation system in further embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, right with reference to the accompanying drawings and embodiments
The more satellite fusion precipitation methods of the present invention and system are further elaborated.It should be appreciated that specific reality described herein
Example is applied only to explain the present invention, is not intended to limit the present invention.
Fig. 1 is the flow diagram of more satellites fusion precipitation method in one embodiment, an implementation as shown in Figure 1
More satellites in example merge precipitation method
Step S10000 obtains sampling periods, and the sampling periods include each year in N preset period of time, Mei Gesuo
The duration of preset period of time is stated less than 1 year, wherein N is positive integer.
Specifically, N is positive integer.The prediction of precipitation is carried out using the method for more satellites fusion precipitation.In the present invention
In provided more satellite fusion precipitation methods, according to the different time precisions of the satellite precipitation data, it can be used for predicting
The precipitation of the different periods such as hour precipitation, intra day ward, monthly total precipitation, it is subsequent specifically to be illustrated using intra day ward
Explanation.Since precipitation has the characteristic of Time Continuous, precipitation of the general antecedent precipitation amount using prediction day to prediction day
It is predicted.Periodic characteristic is presented according to year in precipitation, if the precipitation in 1 year is distributed in spring or summer more, the present invention
The preset period of time refers to the preset period using year as the fixation in period, such as annual March 1 to March 31 this when
Continuous 40 days etc. before section or annual March 1, after having selected 1 year preset period of time, further in selected n-th
It is all fixed in year to select this preset period of time.
Again due to the uncertainty of precipitation, it will appear abnormal fluctuation in the precipitation in certain times, to precipitation
Daily prediction in, can to the data in the Abnormal Precipitation time not calculate, when selecting n-th year, according to need
It asks, can continuously select, can also discontinuously select.
In one embodiment of the invention, for selected prediction day as October 1 was predicted within 2000 when, selection
It is continuous before continuous 40 days before on October 1st, 2000, continuous 40 days before on October 1st, 1999 and on October 1st, 1998
It 40 days, is used as sampling periods within totally 120 days, also may be selected before on October 1st, 1999 continuous 40 days, before on October 1st, 1998
It continuous 40 days and continuous 40 days before on October 1st, 1996, need to be depending on actual conditions.
That is, for needs predict t day 1 year, selection t-x days 1 year to t-1 days 1 year, (n-1)th year the
T-x days to t-1 days (n-1)th year, t-1 days t-x days to the n-th -2 years the n-th -2 years (x is the sample length chosen every year),
Amount to 3x sampling day；Or, choosing t-x days to t-1 days 1 year 1 year, t-x days to the n-th -2 years the n-th -2 years t-1
Day, t-1 days t-x days to the n-th -4 years the n-th -4 years (x is the sample length chosen every year) amount to 3x sampling day.
Step S20000 obtains the precipitation number of the surface precipitation data and at least two satellites in the sampling periods respectively
According to.
Specifically, needing to obtain the drop of surface precipitation data and at least two satellites respectively after sampling periods have been determined
Water number evidence, to carry out satellite fusion precipitation.
The surface precipitation data, are generally from ground station, can be straight as needed have ground station in the region predicted
Connect use, such as do not have ground station, can also be used the ground station of surrounding precipitation data handled after use.Describedly
Face precipitation data is the true value of precipitation.
The selection of the ground station can generate large effect to the calculating of subsequent more satellite fusion precipitation, such as one
In the region that piece needs to predict, according to the geographic properties in region, the precipitation data of representative ground station is selected, or to the greatest extent
The region that amount selection has ground station is predicted, can all make fused precipitation result more accurate.
The precipitation data of the satellite is the retrieving precipitation data of the satellite after primary Calculation.Since difference is defended
The inversion algorithm of star, satellite precision are different, and the error of the retrieving precipitation data of satellite is also by institute estimation range, type of precipitation, drop
The influence of water period, vegetation cover, landform etc. and it is different, need the exclusion as far as possible in subsequent calculating so that fusion
Precipitation result afterwards is more accurate.
Step S30000 determines the prior probability of each satellite.
Specifically, the prior probability of determination each satellite, including according to surface precipitation data and described each
The precipitation data of satellite calculates the prior probability of each satellite, or presets the prior probability of each satellite.In general, described each
The sum of prior probability of satellite is equal to 1.
In actual use, for convenience of calculating, equal prior probability can be preset for each satellite, such as shares 4 and defends
The prior probability of star, every satellite is equal to 1/4=0.25.
The priori of each satellite can also be calculated according to the precipitation data of the surface precipitation data and each satellite
Probability, such as prior probability=(difference of each satellite and surface precipitation data)/each satellite of each satellite and ground drop
The sum of the difference of water number evidence).
Step S40000, according to the prior probability of each satellite, the drop of the surface precipitation data and each satellite
Water number evidence calculates more satellites using probability statistics model and merges precipitation parameter, and more satellite fusion precipitation parameters include described
The fusion weight coefficient of each satellite.
Specifically, the probability statistics model, generally calculates melting for each satellite by iterative calculation using Bayesian model
Close weight coefficient.Using surface precipitation data as true value, after the fusion weight coefficient for determining each satellite, by more satellites
Precipitation data carries out fusion calculation, obtains the fusion precipitation result closest to true value.
More satellites provided by the present embodiment merge precipitation method, according to the difference of sampling periods and the ground station
The calculated result of the difference of selection, the fusion weight coefficient of each satellite is not also identical, and the fusion weight coefficient of each satellite exists
Time and dynamic characteristic is spatially presented, excludes every satellite to the greatest extent by monitoring time or the different institutes in space
The error of generation, so that fused precipitation result is more accurate.
Fig. 2 is the flow diagram that more satellites in another embodiment merge precipitation method, as shown in Figure 2 another
More satellites in embodiment merge precipitation method, are the subdivided steps of above-mentioned steps S40000, comprising:
Step S41000, according to the posteriority point of the posterior probability of the precipitation data of each satellite and more satellites fusion precipitation
Cloth determines the probability density function of more satellite fusion precipitation, wherein the posterior probability of the precipitation data of each satellite is respectively to defend
Posterior probability of the precipitation data of star under the surface precipitation data qualification, the Posterior distrbutionp that more satellites merge precipitation are
The Posterior distrbutionp of more satellites fusion precipitation under precipitation data and the surface precipitation data qualification based on each satellite.
Specifically, the probability density function of the satellite fusion precipitation determined is expressed as follows with formula (1):
wk=p (fk|D)；Formula (2)
Wherein, y is fused precipitation, and D=[y1obs, y2obs ..., yTobs] is the ground precipitation number of sampling periods T
It is the precipitation data of different satellites (or radar) according to, fk=[f1, f2 ..., fK], K is the number of satellite, p (fk | D) it is respectively to defend
Posterior probability of the precipitation data of star under the surface precipitation data qualification, and P (y | fk, D) it is the drop based on each satellite
The Posterior distrbutionp of more satellites fusion precipitation under water number evidence and the surface precipitation data qualification, and p (y | D) it is based on the ground
The probability of more satellites fusion precipitation under the conditions of precipitation data.
Specifically, formula (1) need to meet the condition of formula (2) and formula (3) setting.
The probability density function of more satellite fusion precipitation is iterated calculating, determined described more by step S42000
Satellite merges precipitation parameter.
Specifically, relevant parameter information (such as weight coefficient), the fusion results due to Bayes's multiple-model integration cannot
It immediately arrives at, the probability density function of above-mentioned more satellite fusion precipitation in actual operation, needs to acquire most by continuous iteration
Excellent solution.P (the f obtained after iterative calculationk| D), i.e., the precipitation data of each satellite is under the surface precipitation data qualification
Posterior probability, the fusion weight coefficient of as described each satellite.
More satellites provided by the present embodiment merge precipitation method, provide the general of satellite fusion precipitation using Bayesian model
Rate Formula of density function determines that posteriority of the precipitation data of each satellite under surface precipitation data qualification is general by iterative calculation
Rate, so that it is determined that the fusion weight coefficient of each satellite.The present embodiment using surface precipitation data as true value, comprehensively consider data it
Between internal association, the difference of sampling periods and surface precipitation data at any time, the fusion weight coefficient of each satellite provided is not yet
Together, fused precipitation is made close to true value, to improve the accuracy of more satellite fusion precipitation results to the greatest extent.
Fig. 3 is the flow diagram of more satellites fusion precipitation method in another embodiment, more satellites as shown in Figure 3
Merging precipitation method includes:
The precipitation data of the surface precipitation data and each satellite is carried out normal distribution respectively and turned by step S42100
It changes, obtains the normal distribution precipitation data of ground normal distribution precipitation data and each satellite.
Specifically, when being iterated calculating to formula (1), being mostly used most since precipitation data does not meet normal distribution
Big Expectation Algorithm, and using the premise of EM algorithm is its data fit normal distribution.Therefore, it is necessary to by surface precipitation number
Normal distribution conversion is carried out respectively according to the precipitation data with each satellite.
The present embodiment uses Box-Cox conversion regime, using formula (4) to the precipitation number of ground precipitation data and each satellite
According to progress normal distribution conversion.
Wherein, τ is constant, and y is fused precipitation, ZtFused precipitation after turning for normal state.
Step S42200, according to the ground normal distribution precipitation data, the normal distribution precipitation data of each satellite,
The probability density function of more satellite fusion precipitation is iterated calculating using EM algorithm, it is each after determining optimization
The posterior probability of the precipitation data of satellite.
Specifically, the precipitation data of surface precipitation data and each satellite after normal distribution is converted is utilized maximum
Expectation Algorithm is iterated calculating.The calculating Bayes that the algorithm is built upon the hypothesis of the equal Normal Distribution of K model is more
The most effectual way of model integrated parameter.
The posterior probability of the precipitation data of each satellite after the optimization is determined as each satellite by step S42300
Merge weight coefficient.
More satellites provided by the present embodiment merge precipitation method, provide the general of satellite fusion precipitation using Bayesian model
Rate Formula of density function utilizes the maximum phase after the precipitation data of surface precipitation data and each satellite is carried out normal distribution conversion
Algorithm is hoped to calculate the fusion weight coefficient of each satellite.When calculating data, in conjunction with the data of used algorithm progress
Processing, can make fused precipitation to the greatest extent close to true value, to improve the accurate of more satellite fusion precipitation results
Property.
Fig. 4 is the flow diagram of more satellites fusion precipitation method in further embodiment, more satellites as shown in Figure 4
Merging precipitation method includes:
Step S42210 calculates each satellite according to the precipitation data of each satellite and the surface precipitation data
Precipitation data error.
Specifically, calculating σ2, σ2For the variance of the opposite ground precipitation data of precipitation data of each satellite.
Step S42220, the posteriority of the error of the precipitation data of each satellite, the precipitation data of each satellite is general
Rate, the initial parameter sets to be asked being determined as in the probability density function of more satellite fusion precipitation.
Specifically, with θ={ wk, σ2, k=1,2.., K } and it indicates in the probability density function of more satellite fusion precipitation
Parameter sets to be asked, wherein σ2For the error of the precipitation data of each satellite, WkAfter precipitation data for each satellite
Test probability.
Step S42230, according to the set of the parameter initially to be asked, the ground normal distribution precipitation data and described
The normal distribution precipitation data of each satellite determines the log-likelihood function of the parameter sets initially to be asked.
Specifically, the log-likelihood function of the parameter sets to be asked is formula (5),
Wherein, fkFor the precipitation data of different satellites, wkFor the posterior probability of the precipitation data of each satellite,) expression mean value be fk, variance isNormal distribution.
Step S42240 iterates to calculate the log-likelihood function and determines the maximum likelihood of the log-likelihood function
Value determines the parameter sets to be asked after optimization according to the maximum likelihood value, and the parameter sets to be asked after the optimization include excellent
The posterior probability of the precipitation data of each satellite after change.
Specifically, be difficult to acquire the analytic solutions of θ hardly possible by formula (5), and expectation-maximization algorithm can by expectation and most
Two step of bigization iterates until convergence, obtains maximum likelihood value, to obtain the numerical solution of θ.
More satellites provided by the present embodiment merge precipitation method, furthermore present the probability density of satellite fusion precipitation
The calculating process of function formula, used normal distribution conversion and EM algorithm can guarantee the meter for merging precipitation
It is more accurate to calculate result.
Fig. 5 is the flow diagram of more satellites fusion precipitation method in further embodiment, more satellites as shown in Figure 5
Merging precipitation method includes:
Step S42241 the surface precipitation data and described is respectively defended according to each satellite preset initial weight value
The precipitation data of star calculates the initial variance of the precipitation data of each satellite.
Specifically, firstly, setting primary iteration number i=1, the initial weight value of each satellite are wk=1/k, each satellite
The initial variance of precipitation data is formula (6):
Step S42242 calculates the first of the log-likelihood function according to the initial weight value and the initial variance
Beginning likelihood value.
Specifically, calculating initial likelihood value according to above-mentioned formula (5).
Step S42243 is determined and is hidden according to the precipitation data of the satellite, the variance and the surface precipitation data
Variable.
Specifically, introducing a hidden variable, formula (7) is determinedFor hidden variable.
The hidden variable is iterated calculating, the hidden variable after obtaining iteration by step S42244.
Specifically, by the number of iterations i+1, after calculating iteration according to formula (7)
Step S42245 calculates the iteration weighted value and iteration variance of each satellite according to the hidden variable after the iteration.
Specifically, indicating the iteration weighted value of each satellite with formula (8), the iteration of each satellite is indicated with formula (9)
Variance.
Step S42246 calculates the likelihood value of the log-likelihood function according to the iteration weighted value and iteration variance.
Specifically, iteration weighted value and iteration variance are substituted into formula (5), the likelihood value of log-likelihood function is calculated.
Step S42247 determines the maximum likelihood of the log-likelihood function according to the likelihood value and preset threshold value
Value.
Specifically, convergence is examined, if l (θ) by continuous iterationi-l(θ)i-1Less than or equal to preset threshold value, then
Determine that the likelihood value after current iteration calculates is the maximum likelihood value of the log-likelihood function.
More satellites that the present embodiment provides merge precipitation method, give specifically according to ground precipitation data and to satellite
Precipitation data carry out log-likelihood function maximum likelihood value calculating step, when passing through iterative calculation so that fusion drop
The calculated result of water is more accurate.
Fig. 6 is the flow diagram of more satellites fusion precipitation method in further embodiment, as shown in FIG. 6, defends more
Star merges precipitation method
Step S42210 calculates each satellite according to the precipitation data of each satellite and the surface precipitation data
Precipitation data error.
Step S42220, the posteriority of the error of the precipitation data of each satellite, the precipitation data of each satellite is general
Rate, the initial parameter sets to be asked being determined as in the probability density function of more satellite fusion precipitation.
Step S42230, according to the set of the parameter initially to be asked, the ground normal distribution precipitation data and described
The normal distribution precipitation data of each satellite determines the log-likelihood function of the parameter sets initially to be asked.
Step S42240 iterates to calculate the log-likelihood function and determines the maximum likelihood of the log-likelihood function
Value determines the parameter sets to be asked after optimization according to the maximum likelihood value, and the parameter sets to be asked after the optimization include excellent
The posterior probability of the precipitation data of each satellite after change.
Step S42250 calculates the probability of more satellite fusion precipitation according to the parameter sets to be asked after the optimization
The variance of density function；The variance of the probability density function of more satellite fusion precipitation is determined as more satellite fusion drops
The control information of water.
Specifically, more satellites provided by the present embodiment merge precipitation parameter, it further include more satellite fusion precipitation
Control information.
The mean value of the probability density function of more satellite fusion precipitation provided by the present embodiment is formula (10), and variance is public affairs
Formula (11).
Bayes's multiple-model integration the result is that multiple single models continue to optimize after result of weighted average, the mean value
For fused precipitation, the variance is the control information that more satellites merge precipitation.
More satellites provided by the present embodiment merge precipitation method, by providing the variance of fusion precipitation, that is, merge precipitation
Control information, the probability statistics for giving the accuracy of judgement fusion precipitation melt as a result, helping to further increase more satellites
Close the accuracy of precipitation.
The following are the structural schematic diagram of more satellite fusion precipitation systems provided by the present invention, filled provided by each embodiment
Setting is device corresponding to sending method provided by the present invention, each embodiment corresponding to heretofore described method it is detailed
Thin description content is suitable for corresponding device accordingly, repeats no more.
Fig. 7 is the structural schematic diagram of more satellites fusion precipitation system in one embodiment, and more satellites as shown in Figure 7 melt
Closing precipitation system includes:
Period sampling module 10000, for obtaining sampling periods, the sampling periods include each year default in N
Period, the duration of each preset period of time was less than 1 year；
Precipitation data obtains module 20000, for obtaining surface precipitation data in the sampling periods and at least respectively
The precipitation data of two satellites；
Prior probability determining module 30000, for determining the prior probability of each satellite；
Precipitation parameter computing module 40000, for according to the prior probability of each satellite, the surface precipitation data and
The precipitation data of each satellite calculates more satellites using probability statistics model and merges precipitation parameter, more satellite fusion drops
Water parameter includes the fusion weight coefficient of each satellite.
More satellites provided by the present embodiment merge precipitation system, according to the difference of sampling periods and the ground station
The calculated result of the difference of selection, the fusion weight coefficient of each satellite is not also identical, and the fusion weight coefficient of each satellite exists
Time and dynamic characteristic is spatially presented, excludes every satellite to the greatest extent by monitoring time or the different institutes in space
The error of generation, so that fused precipitation result is more accurate.
Fig. 8 is the structural schematic diagram of more satellites fusion precipitation system in another embodiment, more satellites as shown in Figure 8
Fusion precipitation system structure include:
Probability density function determination unit 41000, for according to the posterior probability of the precipitation data of each satellite and more
Satellite merges the Posterior distrbutionp of precipitation, determines the probability density function of more satellite fusion precipitation, wherein the precipitation of each satellite
The posterior probability of data is posterior probability of the precipitation data of each satellite under the surface precipitation data qualification, more satellites
The Posterior distrbutionp of precipitation is merged as more satellites under the precipitation data based on each satellite and the surface precipitation data qualification
Merge the Posterior distrbutionp of precipitation；
Precipitation parameter computing unit 42000 is merged, for carrying out the probability density function of more satellite fusion precipitation
Iterative calculation determines more satellite fusion precipitation parameters.
More satellites provided by the present embodiment merge precipitation system, provide the general of satellite fusion precipitation using Bayesian model
Rate Formula of density function determines that posteriority of the precipitation data of each satellite under surface precipitation data qualification is general by iterative calculation
Rate, so that it is determined that the fusion weight coefficient of each satellite.The present embodiment using surface precipitation data as true value, comprehensively consider data it
Between internal association, the difference of sampling periods and surface precipitation data at any time, the fusion weight coefficient of each satellite provided is not yet
Together, fused precipitation is made close to true value, to improve the accuracy of more satellite fusion precipitation results to the greatest extent.
Fig. 9 is the structural schematic diagram of more satellites fusion precipitation system in another embodiment, more satellites as shown in Figure 9
Fusion precipitation system structure include:
Normal distribution conversion subunit 42100, for by the precipitation data of the surface precipitation data and each satellite
Normal distribution conversion is carried out respectively, obtains the normal distribution precipitation data of ground normal distribution precipitation data and each satellite；
Greatest hope computation subunit 42200, for according to the ground normal distribution precipitation data, each satellite
The probability density function of more satellite fusion precipitation is iterated meter using EM algorithm by normal distribution precipitation data
It calculates, the posterior probability of the precipitation data of each satellite after determining optimization；And by the precipitation data of each satellite after the optimization
Posterior probability is determined as the fusion weight coefficient of each satellite.More satellites provided by the present embodiment merge precipitation system, benefit
The formula of probability density function that satellite fusion precipitation is provided with Bayesian model, by the precipitation number of surface precipitation data and each satellite
After carrying out normal distribution conversion, the fusion weight coefficient of each satellite is calculated using EM algorithm.It is counted to data
When calculation, in conjunction with the data processing that used algorithm carries out, fused precipitation can be made to the greatest extent close to true value, from
And improve the accuracy of more satellite fusion precipitation results.
Figure 10 is the structural schematic diagram of more satellites fusion precipitation system in further embodiment, and as shown in Figure 10 defend more
Star fusion precipitation system structure include:
Error computation component 42210, for according to each satellite precipitation data and the surface precipitation data, calculate
The error of the precipitation data of each satellite；
Initially parameter obtaining component 42220 to be asked, for by the error of the precipitation data of each satellite, each satellite
Precipitation data posterior probability, the initial parameter set to be asked being determined as in the probability density function of the more satellites fusion precipitation
It closes；
Likelihood function securing component 42230, for the set according to the parameter initially to be asked, the ground normal distribution
The normal distribution precipitation data of precipitation data and each satellite determines the log-likelihood letter of the parameter sets initially to be asked
Number；
Maximum likelihood value securing component 42240, for iterating to calculate the log-likelihood function and determining the logarithm seemingly
The maximum likelihood value of right function determines the parameter sets to be asked after optimization according to the maximum likelihood value, after the optimization to
Seeking parameter sets includes the posterior probability of the precipitation data of each satellite after optimization.Specifically for preset according to each satellite
Initial weight value, the precipitation data of the surface precipitation data and each satellite calculate the precipitation data of each satellite
Initial variance；According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；Root
According to the precipitation data of the satellite, the variance and the surface precipitation data, hidden variable is determined；By the hidden variable into
Row iteration calculates, the hidden variable after obtaining iteration；By the hidden variable after the iteration, the iteration weighted value of each satellite is calculated
With iteration variance；According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；According to described
Likelihood value and preset threshold value determine the maximum likelihood value of the log-likelihood function.
More satellites provided by the present embodiment merge precipitation system, furthermore present the probability density of satellite fusion precipitation
The calculating process of function formula, used normal distribution conversion and EM algorithm can guarantee the meter for merging precipitation
It is more accurate to calculate result.
Figure 11 is the structural schematic diagram of more satellites fusion precipitation system in further embodiment, and as shown in Figure 10 defend more
Star fusion precipitation system structure include:
Error computation component 42210, for according to each satellite precipitation data and the surface precipitation data, calculate
The error of the precipitation data of each satellite；
Initially parameter obtaining component 42220 to be asked, for by the error of the precipitation data of each satellite, each satellite
Precipitation data posterior probability, the initial parameter set to be asked being determined as in the probability density function of the more satellites fusion precipitation
It closes；
Likelihood function securing component 42230, for the set according to the parameter initially to be asked, the ground normal distribution
The normal distribution precipitation data of precipitation data and each satellite determines the log-likelihood letter of the parameter sets initially to be asked
Number；
Maximum likelihood value securing component 42240, for iterating to calculate the log-likelihood function and determining the logarithm seemingly
The maximum likelihood value of right function determines the parameter sets to be asked after optimization according to the maximum likelihood value, after the optimization to
Seeking parameter sets includes the posterior probability of the precipitation data of each satellite after optimization.Specifically for preset according to each satellite
Initial weight value, the precipitation data of the surface precipitation data and each satellite calculate the precipitation data of each satellite
Initial variance；According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；Root
According to the precipitation data of the satellite, the variance and the surface precipitation data, hidden variable is determined；By the hidden variable into
Row iteration calculates, the hidden variable after obtaining iteration；By the hidden variable after the iteration, the iteration weighted value of each satellite is calculated
With iteration variance；According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；According to described
Likelihood value and preset threshold value determine the maximum likelihood value of the log-likelihood function.
Variance computation module 42250, for calculating more satellite fusions according to the parameter sets to be asked after the optimization
The variance of the probability density function of precipitation；
Control information determines component 42260, for the variance of the probability density function of more satellite fusion precipitation is true
It is set to the control information of more satellite fusion precipitation.
More satellites provided by the present embodiment merge precipitation system, by providing the variance of fusion precipitation, that is, merge precipitation
Control information, the probability statistics for giving the accuracy of judgement fusion precipitation melt as a result, helping to further increase more satellites
Close the accuracy of precipitation.
Further, it the present invention also provides a kind of verification method of more satellite fusion precipitation methods, is mentioned based on the present invention
The more satellites supplied merge precipitation method, and specific evaluation index of verifying includes absolute relative error (RBE), root-mean-square error
(RMSE), Pearson correlation coefficient (CC), standard deviation (SD), detectivity (POD), Euclid approach degree (e).
Wherein, absolute relative error (RBE), root-mean-square error (RMSE), standard deviation (SD) describe the ground of ground station
Error and deviation between data and satellite fusion precipitation.Pearson correlation coefficient (CC) for describe ground station observation with
Satellite merges the degree of fitting between precipitation.Euclid approach degree (e) is a kind of measurement of degree of closeness between two fuzzy subsets,
Above-mentioned five other indexs can be combined, be the overall target for the description satellite estimation effect that can be quantified.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (14)
1. a kind of more satellites merge precipitation method, which is characterized in that the described method includes:
Sampling periods are obtained, the sampling periods include each year in N preset period of time, the duration of each preset period of time
Less than 1 year, wherein N was positive integer；
The precipitation data of the surface precipitation data and at least two satellites in the sampling periods is obtained respectively；
Determine the prior probability of each satellite；
According to the prior probability of each satellite, the precipitation data of the surface precipitation data and each satellite, probability is utilized
Statistical model calculates more satellites and merges precipitation parameter, and more satellite fusion precipitation parameters include the fusion weight of each satellite
Coefficient.
2. more satellites according to claim 1 merge precipitation method, which is characterized in that the elder generation of determination each satellite
Test probability, comprising:
The prior probability of each satellite is calculated according to the precipitation data of the surface precipitation data and each satellite, or
Obtain the preset prior probability of each satellite.
3. more satellites according to claim 1 merge precipitation method, which is characterized in that the elder generation according to each satellite
The precipitation data for testing probability, the surface precipitation data and each satellite calculates more satellites using probability statistics model and merges
Precipitation parameter, comprising:
According to the Posterior distrbutionp of the posterior probability of the precipitation data of each satellite and more satellites fusion precipitation, determine that more satellites melt
Close the probability density function of precipitation, wherein the posterior probability of the precipitation data of each satellite is that the precipitation data of each satellite exists
The Posterior distrbutionp of posterior probability under the surface precipitation data qualification, more satellite fusion precipitation is based on each satellite
Precipitation data and the surface precipitation data qualification under more satellites fusion precipitation Posterior distrbutionp；
The probability density function of more satellite fusion precipitation is iterated calculating, determines more satellite fusion precipitation ginsengs
Number.
4. more satellites according to claim 3 merge precipitation method, which is characterized in that described by more satellite fusion drops
The probability density function of water is iterated calculating, determines more satellite fusion precipitation parameters, comprising:
The precipitation data of the surface precipitation data and each satellite is subjected to normal distribution conversion respectively, obtains ground normal state
It is distributed the normal distribution precipitation data of precipitation data and each satellite；
According to the normal distribution precipitation data of the ground normal distribution precipitation data and each satellite, more satellites are melted
The probability density function for closing precipitation is iterated calculating using EM algorithm, the precipitation data of each satellite after determining optimization
Posterior probability；
The posterior probability of the precipitation data of each satellite after the optimization is determined as to the fusion weight coefficient of each satellite.
5. more satellites according to claim 4 merge precipitation method, which is characterized in that described according to the ground normal state point
The normal distribution precipitation data of cloth precipitation data and each satellite, by the probability density function benefit of more satellite fusion precipitation
It is iterated calculating with EM algorithm, the posterior probability of the precipitation data of each satellite after determining optimization, comprising:
According to the precipitation data of each satellite and the surface precipitation data, the mistake of the precipitation data of each satellite is calculated
Difference；
By the posterior probability of the error of the precipitation data of each satellite and the precipitation data of each satellite, it is determined as described more
Satellite merges the initial parameter sets to be asked in the probability density function of precipitation；
According to the set of the parameter initially to be asked, the normal distribution of the ground normal distribution precipitation data and each satellite
Precipitation data determines the log-likelihood function of the parameter sets initially to be asked；
It iterates to calculate the log-likelihood function and determines the maximum likelihood value of the log-likelihood function, seemingly according to the maximum
Right value determines the parameter sets to be asked after optimization, and after the optimization includes the precipitation of each satellite after optimizing wait seek parameter sets
The posterior probability of data.
6. more satellites according to claim 5 merge precipitation method, which is characterized in that the iterative calculation logarithm is seemingly
Right function and the maximum likelihood value for determining the log-likelihood function, specifically include:
According to the preset initial weight value of each satellite, the precipitation data of the surface precipitation data and each satellite calculates institute
State the initial variance of the precipitation data of each satellite；
According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；
It is dropped according to the initial likelihood value of the log-likelihood function, the precipitation data of the satellite, the variance and the ground
Water number evidence, determines hidden variable；
The hidden variable is iterated calculating, the hidden variable after obtaining iteration；
According to the hidden variable after the iteration, the iteration weighted value and iteration variance of each satellite are calculated；
According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；
According to the likelihood value of the log-likelihood function and preset threshold value, the maximum likelihood of the log-likelihood function is determined
Value.
7. more satellites according to claim 6 merge precipitation method, which is characterized in that more satellites merge precipitation parameter
It further include the control information of more satellite fusion precipitation, the calculation method of the control information includes:
According to the parameter sets to be asked after the optimization, the variance of the probability density function of more satellite fusion precipitation is calculated；
The variance of the probability density function of more satellite fusion precipitation is determined as to the error letter of more satellite fusion precipitation
Breath.
8. a kind of more satellites merge precipitation system characterized by comprising
Period sampling module, for obtaining sampling periods, the sampling periods include each year in N preset period of time, each
The duration of the preset period of time was less than 1 year, and wherein N is positive integer；
Precipitation data obtains module, for obtaining surface precipitation data in the sampling periods and at least two satellites respectively
Precipitation data；
Prior probability determining module, for determining the prior probability of each satellite；
Precipitation parameter computing module, for described defending according to the prior probability of each satellite, the surface precipitation data and respectively
The precipitation data of star calculates more satellites using probability statistics model and merges precipitation parameter, and more satellites merge precipitation parameter packet
Include the fusion weight coefficient of each satellite.
9. more satellites according to claim 8 merge precipitation system, which is characterized in that the prior probability determining module,
For calculating the prior probability of each satellite, or acquisition according to the precipitation data of the surface precipitation data and each satellite
The preset prior probability of each satellite.
10. more satellites according to claim 8 merge precipitation system, which is characterized in that the precipitation parameter computing module,
Include:
Probability density function determination unit, for according to the posterior probability of the precipitation data of each satellite and more satellites fusion drop
The Posterior distrbutionp of water determines the probability density function of more satellite fusion precipitation, wherein the posteriority of the precipitation data of each satellite
Probability is posterior probability of the precipitation data of each satellite under the surface precipitation data qualification, more satellite fusion precipitation
Posterior distrbutionp is that more satellites under the precipitation data based on each satellite and the surface precipitation data qualification merge precipitation
Posterior distrbutionp；
Precipitation parameter computing unit is merged, for the probability density function of more satellite fusion precipitation to be iterated calculating,
Determine more satellite fusion precipitation parameters.
11. more satellites according to claim 10 merge precipitation system, which is characterized in that the fusion precipitation parameter calculates
Unit, comprising:
Normal distribution conversion subunit, for carrying out just the precipitation data of the surface precipitation data and each satellite respectively
State distribution conversion, obtains the normal distribution precipitation data of ground normal distribution precipitation data and each satellite；
Greatest hope computation subunit, for the normal distribution according to the ground normal distribution precipitation data and each satellite
The probability density function of more satellite fusion precipitation is iterated calculating using EM algorithm, determined by precipitation data
The posterior probability of the precipitation data of each satellite after optimization；And by the posterior probability of the precipitation data of each satellite after the optimization
It is determined as the fusion weight coefficient of each satellite.
12. more satellites according to claim 11 merge precipitation system, which is characterized in that it is single that the greatest hope calculates son
Member, comprising:
Error computation component, for according to each satellite precipitation data and the surface precipitation data, calculate and described respectively defend
The error of the precipitation data of star；
Initially parameter obtaining component to be asked, for by the precipitation number of the error of the precipitation data of each satellite and each satellite
According to posterior probability, the initial parameter sets to be asked being determined as in the probability density function of the more satellites fusion precipitation；
Likelihood function securing component, for the set according to the parameter initially to be asked, the ground normal distribution precipitation data
With the normal distribution precipitation data of each satellite, the log-likelihood function of the parameter sets initially to be asked is determined；
Maximum likelihood value securing component, for iterating to calculate the log-likelihood function and determining the log-likelihood function most
Maximum-likelihood value determines the parameter sets to be asked after optimization according to the maximum likelihood value, the parameter sets to be asked after the optimization
The posterior probability of precipitation data including each satellite after optimization.
13. more satellites according to claim 12 merge precipitation system, which is characterized in that the maximum likelihood value acquisition group
Part is specifically used for:
According to the preset initial weight value of each satellite, the precipitation data of the surface precipitation data and each satellite calculates institute
State the initial variance of the precipitation data of each satellite；
According to the initial weight value and the initial variance, the initial likelihood value of the log-likelihood function is calculated；
It is dropped according to the initial likelihood value of the log-likelihood function, the precipitation data of the satellite, the variance and the ground
Water number evidence, determines hidden variable；
The hidden variable is iterated calculating, the hidden variable after obtaining iteration；
According to the hidden variable after the iteration, the iteration weighted value and iteration variance of each satellite are calculated；
According to the iteration weighted value and iteration variance, the likelihood value of the log-likelihood function is calculated；
According to the likelihood value of the log-likelihood function and preset threshold value, the maximum likelihood of the log-likelihood function is determined
Value.
14. more satellites according to claim 13 merge precipitation system, which is characterized in that it is single that the greatest hope calculates son
Member, further includes:
Variance computation module, for calculating the general of more satellite fusion precipitation according to the parameter sets to be asked after the optimization
The variance of rate density function；
Control information determines component, described more for the variance of the probability density function of more satellite fusion precipitation to be determined as
The control information of satellite fusion precipitation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610881113.6A CN107918166B (en) | 2016-10-09 | 2016-10-09 | More satellite fusion precipitation methods and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610881113.6A CN107918166B (en) | 2016-10-09 | 2016-10-09 | More satellite fusion precipitation methods and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107918166A CN107918166A (en) | 2018-04-17 |
| CN107918166B true CN107918166B (en) | 2019-10-18 |
Family
ID=61891607
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610881113.6A Active CN107918166B (en) | 2016-10-09 | 2016-10-09 | More satellite fusion precipitation methods and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107918166B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108874734B (en) * | 2018-04-25 | 2022-03-08 | 中国科学院国家空间科学中心 | Global land rainfall inversion method |
| CN108761574B (en) * | 2018-05-07 | 2021-04-27 | 中国电建集团北京勘测设计研究院有限公司 | Rainfall estimation method based on multi-source information fusion |
| 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 |
| CN109993372B (en) * | 2019-04-12 | 2022-11-22 | 淮河水利委员会水文局(信息中心) | Flood probability forecasting method based on multi-source uncertainty |
| 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 |
| CN111308581B (en) * | 2020-04-10 | 2021-10-22 | 海南省气象科学研究所 | Radar-rain gauge combined rainfall estimation method based on space-time local model |
| CN112861072B (en) * | 2021-02-09 | 2021-10-19 | 河海大学 | Satellite-ground multi-source rainfall self-adaptive dynamic fusion method |
| CN113159378B (en) * | 2021-03-15 | 2022-04-05 | 中国科学院地理科学与资源研究所 | Rainfall estimation method combining mountain runoff without survey stations and remote sensing rainfall |
| CN114037017B (en) * | 2021-11-25 | 2022-10-21 | 西安电子科技大学 | Data fusion method based on error distribution fitting |
| WO2024036516A1 (en) * | 2022-08-17 | 2024-02-22 | 中山大学 | Precipitation normalization analysis method and system based on gradient parameter optimization |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101349767A (en) * | 2008-09-05 | 2009-01-21 | 国家卫星气象中心 | High resolution precipitation data processing method |
| CN103942941A (en) * | 2014-04-11 | 2014-07-23 | 中国人民解放军61139部队 | Mobile monitoring fusion platform based on geographic information system (GIS) |
| CN105068153A (en) * | 2015-06-19 | 2015-11-18 | 中国气象科学研究院 | Regional automatic rainfall station hourly rainfall data quality control system and method |
| CN105608840A (en) * | 2016-03-09 | 2016-05-25 | 长江水利委员会水文局 | Mountain torrents early warning platform based on fused quantitative rainfall forecast algorithm, and early warning method thereof |
-
2016
- 2016-10-09 CN CN201610881113.6A patent/CN107918166B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101349767A (en) * | 2008-09-05 | 2009-01-21 | 国家卫星气象中心 | High resolution precipitation data processing method |
| CN103942941A (en) * | 2014-04-11 | 2014-07-23 | 中国人民解放军61139部队 | Mobile monitoring fusion platform based on geographic information system (GIS) |
| CN105068153A (en) * | 2015-06-19 | 2015-11-18 | 中国气象科学研究院 | Regional automatic rainfall station hourly rainfall data quality control system and method |
| CN105608840A (en) * | 2016-03-09 | 2016-05-25 | 长江水利委员会水文局 | Mountain torrents early warning platform based on fused quantitative rainfall forecast algorithm, and early warning method thereof |
Non-Patent Citations (9)
| Title |
|---|
| A conceptual model for constructing high‐resolution gauge‐satellite merged precipitation analyses;Pingping Xie 等;《JOURNAL OF GEOPHYSICAL RESEARCH》;20111108;第116卷;第D21106页 * |
| A high spatiotemporal gauge‐satellite merged precipitation analysis over China;Yan Shen 等;《Journal of Geophysical Research: Atmospheres》;20140324(第119期);第3063-3075页 * |
| Improvement of Multi-Satellite Real-Time Precipitation Products for Ensemble Streamflow Simulation in a Middle Latitude Basin in South China;Shanhu Jiang 等;《Water Resour Manage》;20140430;第28卷(第2014期);第2259-2278页 * |
| The Global Precipitation [8] Climatology Project (GPCP) combined precipitation dataset;George J. Huffman 等;《Bulletin of the AmericanMeteorological Society》;19970131;第78卷(第1期);第5-20页 * |
| Variational merged of hourly gauge‐satellite precipitation in China_ Preliminary results;Huan Li 等;《Journal of Geophysical Research: Atmospheres》;20151005(第120期);第9897-9905页 * |
| 基于多源信息的降水空间估计及其水文应用研究;胡庆芳;《中国优秀博士学位论文全文数据库基础科学辑》;20140715(第7期);第A012-7页 * |
| 基于贝叶斯融合方法的高分辨率地面-卫星-雷达三源降水融合试验;潘旸 等;《气象学报》;20151230;第73卷(第1期);第177-186页 * |
| 多平台(雷达、卫星、雨量计)降水信息的融合技术初探;高晓荣 等;《高原气象》;20130430;第32卷(第2期);第549-555页 * |
| 多源降雨观测与融合及其在长江流域的水文应用;李哲;《中国优秀博士学位论文全文数据库基础科学辑》;20160715(第7期);第A009-1页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107918166A (en) | 2018-04-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107918165B (en) | More satellites based on space interpolation merge Prediction of Precipitation method and system | |
| CN107918166B (en) | More satellite fusion precipitation methods and system | |
| Sun et al. | A review of global precipitation data sets: Data sources, estimation, and intercomparisons | |
| Gebregiorgis et al. | To what extent is the day 1 GPM IMERG satellite precipitation estimate improved as compared to TRMM TMPA‐RT? | |
| Chen et al. | Evaluation of TRMM 3B42 precipitation estimates of tropical cyclone rainfall using PACRAIN data | |
| Cordoba et al. | Diagnosing atmospheric motion vector observation errors for an operational high‐resolution data assimilation system | |
| Imhoff et al. | Spatial and temporal evaluation of radar rainfall nowcasting techniques on 1,533 events | |
| Kirstetter et al. | Toward a framework for systematic error modeling of spaceborne precipitation radar with NOAA/NSSL ground radar–based National Mosaic QPE | |
| Demuth et al. | Evaluation of Advanced Microwave Sounding Unit tropical-cyclone intensity and size estimation algorithms | |
| Rabatel et al. | Changes in glacier equilibrium-line altitude in the western Alps from 1984 to 2010: evaluation by remote sensing and modeling of the morpho-topographic and climate controls | |
| Loew et al. | A dynamic approach for evaluating coarse scale satellite soil moisture products | |
| Tuleya et al. | Evaluation of GFDL and simple statistical model rainfall forecasts for US landfalling tropical storms | |
| Aryee et al. | Development of high spatial resolution rainfall data for Ghana | |
| Pereira Filho et al. | Satellite Rainfall Estimates Over South America–Possible Applicability to the Water Management of Large Watersheds 1 | |
| Lockhoff et al. | Evaluation of satellite-retrieved extreme precipitation over Europe using gauge observations | |
| Behrangi et al. | REFAME: Rain estimation using forward-adjusted advection of microwave estimates | |
| Ouma et al. | Multitemporal comparative analysis of TRMM-3B42 satellite-estimated rainfall with surface gauge data at basin scales: daily, decadal and monthly evaluations | |
| Jung et al. | Radar‐based cell tracking with fuzzy logic approach | |
| Lebel et al. | Rainfall estimation in the Sahel: What is the ground truth? | |
| Otsuka et al. | Nowcasting with data assimilation: A case of global satellite mapping of precipitation | |
| Waller et al. | Observation error statistics for Doppler radar radial wind superobservations assimilated into the DWD COSMO-KENDA system | |
| Kober et al. | Aspects of short‐term probabilistic blending in different weather regimes | |
| Borsche et al. | Methodologies to characterize uncertainties in regional reanalyses | |
| Fan et al. | A comparative study of four merging approaches for regional precipitation estimation | |
| Hong et al. | 15 Global Precipitation Estimation and Applications |
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 |