CN113267834A - Fusion rainfall forecasting method based on multi-model integration - Google Patents
Fusion rainfall forecasting method based on multi-model integration Download PDFInfo
- Publication number
- CN113267834A CN113267834A CN202011372595.5A CN202011372595A CN113267834A CN 113267834 A CN113267834 A CN 113267834A CN 202011372595 A CN202011372595 A CN 202011372595A CN 113267834 A CN113267834 A CN 113267834A
- Authority
- CN
- China
- Prior art keywords
- forecast
- precipitation
- fusion
- rainfall
- numerical
- 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.)
- Pending
Links
- 230000004927 fusion Effects 0.000 title claims abstract description 62
- 230000010354 integration Effects 0.000 title claims abstract description 17
- 238000013277 forecasting method Methods 0.000 title claims abstract description 16
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000013459 approach Methods 0.000 claims abstract description 9
- 238000001556 precipitation Methods 0.000 claims description 104
- 238000012937 correction Methods 0.000 claims description 39
- 238000005516 engineering process Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 18
- 230000003287 optical effect Effects 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 10
- 238000013213 extrapolation Methods 0.000 claims description 5
- 238000004800 variational method Methods 0.000 claims description 5
- 230000032683 aging Effects 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 4
- 230000002265 prevention Effects 0.000 abstract description 2
- 230000033001 locomotion Effects 0.000 description 7
- 230000009897 systematic effect Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000005206 flow analysis Methods 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 230000002035 prolonged effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000010183 spectrum analysis Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- 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
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Environmental & Geological Engineering (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Environmental Sciences (AREA)
- Ecology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Atmospheric Sciences (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to the technical field of rainfall forecast, in particular to a fusion rainfall forecast method based on multi-model integration. In the fusion algorithm, the time-by-time rainfall forecast and the corrected model forecast of the approach forecast determine weight factors by a hyperbolic function, and the weight of the numerical model forecast result is gradually increased along with the extension of forecast timeliness. Has the advantages that: compared with the traditional forecasting method or a single neural network forecasting mode, the method integrates two methods, namely 'getting strong points and compensating for weak points', and not only improves the accuracy of numerical forecasting of the falling area of rainfall, but also improves the accuracy of the strength of rainfall forecast nearby. The method is expected to have greater economic benefit and social benefit on accurate weather forecast and weather disaster prevention.
Description
Technical Field
The invention relates to the technical field of rainfall forecast, in particular to a fusion rainfall forecast method based on multi-model integration.
Background
At present, a short-time nowcasting technology based on radar observation and echo recognition, tracking and extrapolation cannot provide high-quality forecast of the development and evolution of a convection weather system for more than 2h, a mesoscale numerical mode has a large defect in short-time forecast of flow scale quantitative precipitation, and the nowcasting and the numerical forecast are fused, so that the most important way and means for providing effective forecast of the convection scale weather system, particularly 0-6 h of convection strong precipitation at present are provided.
The scheme provides a fusion rainfall forecasting method based on multi-model integration, which carries out fusion forecasting on a deep neural network short-threshold forecasting algorithm and a mesoscale numerical forecasting algorithm by distributing weights in real time, overcomes the forecasting timeliness problem of a proximity forecasting system and the 'spin-up' problem of numerical forecasting, improves the quantitative rainfall forecasting effect of O-6 h, and particularly improves the forecasting accuracy of rainfall intensity. The scheme is that firstly, based on Quantitative Precipitation Estimation (QPE) results of radar detection and automatic meteorological station observation, the quantitative precipitation forecast output in a scale numerical mode is subjected to phase correction in a spectrum space, the deviation of numerical forecast and observation is analyzed and calculated, and an additional numerical forecast correction field is derived; secondly, adjusting the numerical prediction precipitation area and intensity in a corresponding time period according to the characteristic that the numerical prediction correction field meets certain time change distribution; and finally, fusing the corrected numerical model quantitative precipitation forecast and the quantitative precipitation forecast based on the prediction approach technology of the deep neural network by using a hyperbolic tangent weight function, wherein the fusion weight dynamically changes according to the typical circulation characteristic. The fused quantitative precipitation forecast shows that the quantitative precipitation forecast mainly depends on an 'extrapolation' approach forecast result in the first 1-2 hours, and then the contribution of the numerical forecast to the fusion result is gradually increased along with the change of the fusion weight until the quantitative precipitation forecast is dominant in 5-6 hours after the fusion.
Disclosure of Invention
The invention aims to solve the defects in the prior art, and provides a fusion rainfall forecasting method based on multi-model integration.
In order to achieve the purpose, the invention adopts the following technical scheme:
designing a fusion rainfall forecasting method based on multi-model integration, wherein the fusion rainfall forecasting method based on multi-model integration comprises the following steps:
s1, firstly, identifying the error of the rainfall forecast falling area and the rainfall intensity in the mesoscale mode, correcting the position error of the rainfall zone forecast in the numerical weather mode by using the 'phase correction' technology, and simultaneously adjusting the rainfall intensity in the mode according to the rainfall observed in the scene. In the fusion algorithm, the time-by-time rainfall forecast and the corrected model forecast of the approach forecast determine weight factors by a hyperbolic function, and the weight of the numerical model forecast result is gradually increased along with the extension of forecast timeliness.
And S2, acquiring fusion test data, wherein the fusion test data comprise a quantitative precipitation estimation calculation scheme, a 0-6 h quantitative precipitation forecast calculation scheme based on an approach prediction technology and a 0-6 h quantitative precipitation forecast based on a numerical mode.
S3: and finally, performing numerical prediction precipitation field correction and test, wherein the numerical prediction precipitation field correction comprises numerical prediction precipitation field phase correction, numerical prediction precipitation field strength correction and fusion weight adjustment.
Preferably, the quantitative precipitation estimation calculation scheme in S2 includes the following specific steps:
(1) the radar quantitative precipitation estimate (grid data) is interpolated onto the automated station site.
(2) And calculating the precipitation observation increment E of each observation station, namely the deviation between the radar quantitative precipitation estimation and the precipitation observation of the automatic station.
(3) And calculating the deviation increment C of the grid points by using a Cressman analysis technology, wherein the deviation increment C represents the weight of the site value in the influence radius to the grid point analysis, d is the distance between the site and the grid point and is smaller than the influence radius R, and R is 10km which is 10 times of the grid distance.
(4) And correcting the radar quantitative precipitation estimation by using the calculated lattice point deviation increment to obtain the quantitative precipitation estimation observed by the fusion radar and the automatic station, wherein the quantitative precipitation estimation is used as a true value of precipitation used in the fusion forecast test.
Preferably, in the calculation scheme of the 0-6 h quantitative precipitation forecast based on the nowcasting technology in S2, the echo forecast time period is prolonged to 6h by adjusting an extrapolation algorithm and parameters on the basis that the original forecast time period is 0-2 h, and then the 0-6 h quantitative precipitation forecast based on the nowcasting technology is calculated by using a local Z.R relation.
Preferably, the 0-6 h quantitative precipitation forecast based on the numerical pattern in S2 is provided by BJ-RUC.
Preferably, the numerical prediction rainfall field phase correction method in S3 adopts two-step correction, and in the first step, a fast fourier transform method is used to ensure that the rain belt overall displacement deviation is corrected; and secondly, reasonably adjusting the trend of the rain belt and the small-range rainfall area by using a multi-scale optical flow variational method, so that the rainfall area forecasted by the BJ-RUC is more consistent with the actual situation.
Preferably, the numerical forecast rainfall field intensity correction in S3 includes a rainfall intensity correction method and a rainfall intensity correction test.
Preferably, the fusion weight adjusting method in S3 is substantially 'making good for the deficiency', the 'deficiency' predicted by the 'long' supplement value of the nowcast is taken at the early stage of fusion, and the 'deficiency' predicted by the 'long' supplement value of the nowcast is taken at the late stage of fusion, so as to achieve a better prediction effect within 0-6 h.
The fusion rainfall forecasting method based on multi-model integration has the beneficial effects that:
1. a digital forecast precipitation phase correction method is provided, so that the trend of a rain belt and a small-range precipitation area are reasonably adjusted, and the forecast precipitation area is more consistent with the actual situation after two steps of precipitation phase adjustment.
2. In the aspect of fusion weight setting, according to the weather system type and the spectral spatial correlation of radar data, the spatio-temporal scales of different precipitation systems are combined, different weights are given to two precipitation fields under different conditions, and hyperbolic tangent weight curves under different conditions are given by an empirical formula. When the numerical prediction and the nowcast scoring effect are good, the fusion method can effectively improve the prediction effect, overcomes the defects of different prediction methods, and improves the level of quantitative rainfall prediction for 0-6 h as a whole.
3. Compared with the traditional forecasting method or a single neural network forecasting mode, the method integrates two methods, namely 'getting strong points and compensating for weak points', and not only improves the accuracy of numerical forecasting of the falling area of rainfall, but also improves the accuracy of the strength of rainfall forecast nearby. The method is expected to have greater economic benefit and social benefit on accurate weather forecast and weather disaster prevention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
A fusion rainfall forecasting method based on multi-model integration comprises the following steps:
s1, firstly, identifying the error of the rainfall forecast falling area and the rainfall intensity in the mesoscale mode, correcting the position error of the rainfall zone forecast in the numerical weather mode by using the 'phase correction' technology, and simultaneously adjusting the rainfall intensity in the mode according to the rainfall observed in the scene. In the fusion algorithm, the time-by-time rainfall forecast and the corrected model forecast of the approach forecast determine weight factors by a hyperbolic function, and the weight of the numerical model forecast result is gradually increased along with the extension of forecast timeliness.
And S2, acquiring fusion test data, wherein the fusion test data comprise a quantitative precipitation estimation calculation scheme, a 0-6 h quantitative precipitation forecast calculation scheme based on an approach prediction technology and a 0-6 h quantitative precipitation forecast based on a numerical mode.
S3: and finally, performing numerical prediction precipitation field correction and test, wherein the numerical prediction precipitation field correction comprises numerical prediction precipitation field phase correction, numerical prediction precipitation field strength correction and fusion weight adjustment.
Example one
The quantitative precipitation estimation calculation scheme in the S2 comprises the following specific steps:
(1) the radar quantitative precipitation estimate (grid data) is interpolated onto the automated station site.
(2) And calculating the precipitation observation increment E of each observation station, namely the deviation between the radar quantitative precipitation estimation and the precipitation observation of the automatic station.
(3) Calculating the variance increment C of the lattice points by using the Cressman analysis technology
In the formula (I), the compound is shown in the specification,
and (3) representing the weight of the site value in the influence radius to the grid point analysis, wherein d is the distance between the site and the grid point and is smaller than the influence radius R, and R is 10km which is 10 times of the grid distance.
(4) And correcting the radar quantitative precipitation estimation by using the calculated lattice point deviation increment to obtain the quantitative precipitation estimation observed by the fusion radar and the automatic station, wherein the quantitative precipitation estimation is used as a true value of precipitation used in the fusion forecast test.
The basic idea for correcting the radar quantitative precipitation estimation is that the precipitation of the automatic station after quality control is considered to be objective and can represent average precipitation of an area, the radar quantitative precipitation estimation can reflect the structure of a precipitation field, but systematic deviation on the quantity value exists, and the radar quantitative precipitation estimation result of the automatic station and the radar quantitative precipitation estimation result of the area around the automatic station can be corrected by utilizing the deviation between the radar quantitative precipitation estimation of the position of the automatic station and the rainfall observation of the automatic station.
Example two
In the 0-6 h quantitative precipitation forecast calculation scheme based on the nowcasting technology in the S2, on the basis that the original forecast aging is 0-2 h, the echo forecast aging is prolonged to 6h by adjusting an 'extrapolation' algorithm and parameters, and then the 0-6 h quantitative precipitation forecast based on the nowcasting technology is calculated by utilizing a local Z.R relation.
EXAMPLE III
The 0-6 h quantitative precipitation forecast based on the numerical pattern in S2 is provided by BJ-RUC;
the BJ-RUC system is a 3h period rapid updating circulation forecasting system established based on a three-dimensional variation assimilation technology and a WRF-ARW mode, the system is 3-nested, the resolution ratio is respectively 27 km, 9 km and 3km, the BJ-RUC system carries out three-dimensional variation assimilation once every 3h, the assimilation data comprises global observation data such as global sounding, ground, ships and airplanes obtained from a real-time database of a meteorological information center, regional automatic station data and foundation global positioning system degradable water volume data, 24h forecasting is carried out after each 3h of assimilation, the output interval of a forecasting product is 1h, and therefore time-by-time mode quantitative rainfall forecasting is obtained.
Example four
The numerical prediction rainfall field phase correction method in the S3 adopts two-step correction, wherein in the first step, a fast Fourier transform method is firstly used to ensure that the whole displacement deviation of the rain belt is corrected; and secondly, reasonably adjusting the trend of the rain belt and the small-range rainfall area by using a multi-scale optical flow variational method, so that the rainfall area forecasted by the BJ-RUC is more consistent with the actual situation.
Fast fourier transform algorithm: the fast Fourier transform algorithm is one of the bases of spectrum analysis, can effectively extract the characteristic quantities of frequency, amplitude, phase angle and the like of signals containing periodic components in a mixed signal, and has the main idea that the signals in a time domain are decomposed into a mode of superposing a plurality of sinusoidal signals with different amplitude phases, so that the spectrum analysis of the signals is realized.
Multi-scale optical flow variation method: the concept of optical flow is derived from an object motion detection method in an image processing technology, and is very useful in computer vision applications, in brief, the motion of an object can be described by a motion vector field, and when the object is projected on a plane image, the motion is often represented by the change of gray scale or luminosity distribution of different points in an image sequence along with time, the change trend of the gray scale or luminosity is defined as an optical flow field and is represented by a vector field (",'), because the optical flow field involves two independent variables" and v, the optical flow equation is not enough to solve the problem, and an additional condition or data must be added to obtain a unique solution, the scheme adopts an algorithm based on a variational method and optical flow field smoothing, writes the optical flow equation as a cost function, and assumes that the optical flow has continuity of physical motion, and adds a smoothing requirement to the vector field (", v), expressed by cost function, the total cost function J is minimized by using variational method,
since the cost function terms J0 and J are different in magnitude and unit, a parameter r needs to be added to adjust the smooth constraint proportion, since there are multiple scale features in objective analysis of rain area movement, optical flow analysis is performed in 7 levels with different resolutions according to relevant radar reflectivity factor data, and the corresponding optical flow fields are solved one by one from low to high resolutions (i.e. large to small scales), specifically, the analysis scale of the 1 st level is the largest, the whole calculated width (600km × 600km) is taken, the 2 nd level is determined as one fifth width of the calculation domain, and the resolution of each layer is set as the double of the previous level from the 3 rd level to the final 7 th level, and in the optical flow analysis of different levels, the required smooth constraint proportion is different due to the different set scales or resolutions. In brief, the value of y increases with the rise of the hierarchy (the 7 levels are sequentially 0.001, O.001, 0.01, 0.05 and 0.1), the smoothing effect becomes more important on a small scale, the motion vector of the rain belt at each set resolution or scale can be obtained by the multi-scale analysis technology, and the overall group velocity and the phase velocity of the individual echo can be reflected in the final analysis field. And obtaining the optimal translation position of the forecast rainfall field by solving the error square minimum value of the forecast rainfall field and the actual rainfall field after translation, and further obtaining the value of the phase correction vector. Once the optimal translation is determined, the same mapping set parameters can be applied to the correction of the next temporal mode quantitative precipitation forecast field with the same initial field.
EXAMPLE five
The numerical forecast rainfall field strength correction in the S3 comprises a rainfall strength correction method and a rainfall strength correction test.
The precipitation intensity correction method comprises the following steps: the precipitation intensity correction is adjusted by approximating the mode quantitative precipitation forecast to the quantitative precipitation estimate, assuming that the mode quantitative precipitation forecast and the quantitative precipitation estimate field satisfy Weber distribution, and the cumulative distribution functions of the two fields are the same, using Weber distribution with 2 parameters in the test, wherein the cumulative distribution function is
Wherein x is more than 0 and is precipitation, a and b are more than 0, wherein b is a shape parameter, a mouth is a scale parameter, the shape parameter is usually [1, 7], and the shape parameter is an important parameter and determines the basic shape of the distribution density curve; the scale parameters play a role in enlarging or reducing the curve, but do not influence the distribution shape, and the parameters a and b of the Weber distribution are obtained by solving through multi-sample operation during each operation. It is worth noting that in different operation of different precipitation cases, weber distribution parameters are different, and the intensity adjustment condition of each time is different, so that the intensity adjustment of each case is effective.
EXAMPLE six
The fusion weight adjusting method in the S3 is essentially 'making strong points and making weak points', the 'long' complementary value forecast 'short' of the nowcasting is taken at the early stage of fusion, the 'long' complementary value forecast 'short' of the nowcasting is taken at the later stage of fusion, and a good forecast effect is achieved within 0-6 h. The weight change of the mode is represented by a hyperbolic tangent, two end points of the tangent are given according to the weather type of precipitation and the weather change experience of a forecaster, and different weights are taken under different conditions by combining the space-time scales of different precipitation systems.
Taking an empirical equation in calculating numerical mode weights
In the formula, a and p are respectively the weights of the numerical modes of z being 0 and 6 (representing the current time and the future 6h), the values of a and lu are determined according to the weather change experience of forecasters, radar climatology, the strength of a convection system and the like, and y represents the middle part W in the fusion period. (f) The slope of the weight curve is adjusted), the value is used for determining the change speed of the weight curve, and the y and mouth values are determined according to the type of a precipitation system, the speed of a precipitation process and the like. For a local strong convection system, the effect of the nowcasting within 1h is better, so the fusion weight within 1h before fusion is kept unchanged, namely the numerical prediction weight within 1h before fusion keeps the threshold value unchanged; the prediction effect of the nowcasting is sharply reduced along with the time extension, and the nowcasting has no reference value after 6h, so that the numerical prediction weight _19 at the 6 th hour is 1.
The fusion of the nowcasting and the numerical mode is realized by adopting a tangent dynamic weight fusion method, the relative weight of the output prediction values of the nowcasting and the numerical prediction needs to be adjusted along with the change of time, the nearest prediction takes the maximum weight in a short prediction time, but the error of the nowcasting increases along with the extension of the prediction time, and a larger weight is given to the numerical prediction result to prolong the prediction time and accuracy in a long prediction time period. The weight change of the mode is represented by a hyperbolic tangent, two end points of the tangent are given according to the weather type of precipitation and the weather change experience of a forecaster, and different weights are taken under different conditions by combining the space-time scales of different precipitation systems.
Taking an empirical equation in calculating numerical mode weights
In the formula, a and p are respectively the weights of the numerical modes of z being 0 and 6 (representing the current time and the future 6h), the values of a and lu are determined according to the weather change experience of forecasters, radar climatology, the strength of a convection system and the like, and y represents the middle part W in the fusion period. (f) The slope of the weight curve is adjusted), the value is used for determining the change speed of the weight curve, and the y and mouth values are determined according to the type of a precipitation system, the speed of a precipitation process and the like. For a local strong convection system, the effect of the nowcasting within 1h is better, so the fusion weight within 1h before fusion is kept unchanged, namely the numerical prediction weight within 1h before fusion keeps the threshold value unchanged; the prediction effect of the nowcasting is sharply reduced along with the time extension, and the nowcasting has no reference value after 6h, so that the numerical prediction weight _19 at the 6 th hour is 1.
The test determines the y and a values according to the weather system type and the spectral-spatial correlation of radar data. When a strong convection system example test with fast rainfall process change is carried out, the current level of the forecasting capacity of numerical forecasting on the convection rainfall process is limited, so that the numerical forecasting weight a at the initial moment is O.05, the spectrum space correlation of radar data is small when the convective weather process change is fast, and the y value is 1.2; in the case of a systematic rainfall experiment in which the rainfall process changes slowly, the numerical prediction has certain prediction capability on the systematic rainfall process, and particularly the numerical prediction after phase and intensity correction has higher application value, so that the numerical prediction weight a at the initial time is taken as 0.1, the systematic weather process changes slowly, the spectral-spatial correlation of radar data is large, and the y value is 0.7. According to the equation, the fusion prediction of the first 1h is mainly based on the result of the proximity prediction, the specific gravity of the numerical mode is correspondingly increased in the later stage of the prediction, the weight of the numerical mode is slowly increased in the early stage of the fusion, the weight is quickly increased in the middle stage, and the weight increase rate of the numerical mode in the later stage tends to be smooth.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (7)
1. A fusion rainfall forecasting method based on multi-model integration is characterized in that: the fusion rainfall forecasting method based on multi-model integration comprises the following steps:
s1, firstly, identifying the error of the rainfall forecast falling area and the rainfall intensity in the mesoscale mode, correcting the position error of the rainfall zone forecast in the numerical weather mode by using the 'phase correction' technology, and simultaneously adjusting the rainfall intensity in the mode according to the rainfall observed in the scene. In the fusion algorithm, the time-by-time rainfall forecast and the corrected model forecast of the approach forecast determine weight factors by a hyperbolic function, and the weight of the numerical model forecast result is gradually increased along with the extension of forecast timeliness.
And S2, acquiring fusion test data, wherein the fusion test data comprise a quantitative precipitation estimation calculation scheme, a 0-6 h quantitative precipitation forecast calculation scheme based on an approach prediction technology and a 0-6 h quantitative precipitation forecast based on a numerical mode.
S3: and finally, performing numerical prediction precipitation field correction and test, wherein the numerical prediction precipitation field correction comprises numerical prediction precipitation field phase correction, numerical prediction precipitation field strength correction and fusion weight adjustment.
2. The fusion precipitation forecasting method based on multi-model integration according to claim 1, wherein the quantitative precipitation estimation calculation scheme in S2 comprises the following specific steps:
(1) the radar quantitative precipitation estimate (grid data) is interpolated onto the automated station site.
(2) And calculating the precipitation observation increment E of each observation station, namely the deviation between the radar quantitative precipitation estimation and the precipitation observation of the automatic station.
(3) And calculating the deviation increment C of the grid points by using a Cressman analysis technology, wherein the deviation increment C represents the weight of the site value in the influence radius to the grid point analysis, d is the distance between the site and the grid point and is smaller than the influence radius R, and R is 10km which is 10 times of the grid distance.
(4) And correcting the radar quantitative precipitation estimation by using the calculated lattice point deviation increment to obtain the quantitative precipitation estimation observed by the fusion radar and the automatic station, wherein the quantitative precipitation estimation is used as a true value of precipitation used in the fusion forecast test.
3. The fusion precipitation forecast method based on multiple model integration as claimed in claim 1, wherein said 0-6 h quantitative precipitation forecast calculation scheme based on the nowcasting technique in S2 is that based on the original forecast aging of 0-2 h, the echo forecast aging is extended to 6h by adjusting the "extrapolation" algorithm and parameters, and then a local Z.R relation is used to calculate the 0-6 h quantitative precipitation forecast based on the nowcasting technique.
4. The method of claim 1, wherein the quantitative 0-6 h precipitation forecast based on numerical model is provided by BJ-RUC in S2.
5. The fusion rainfall forecasting method based on multi-model integration according to claim 1, wherein the numerical forecasting rainfall field phase correction method in S3 adopts two-step correction, and in the first step, a fast fourier transform method is used to ensure that the displacement deviation of the whole rain belt is corrected; and secondly, reasonably adjusting the trend of the rain belt and the small-range rainfall area by using a multi-scale optical flow variational method, so that the rainfall area forecasted by the BJ-RUC is more consistent with the actual situation.
6. The fusion precipitation forecasting method based on multi-model integration according to claim 1, wherein the numerical forecasting precipitation field strength correction in S3 comprises a precipitation strength correction method and a precipitation strength correction test.
7. The fusion rainfall forecasting method based on multi-model integration according to claim 1, wherein the fusion weight adjustment method in S3 is substantially 'make good for short', the 'long' complement value forecasted 'short' in the early stage of fusion is taken from the forecast, and the 'long' complement value forecasted 'short' in the later stage of fusion is taken from the forecast, so as to achieve a better forecasting effect within 0-6 h.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011372595.5A CN113267834A (en) | 2020-11-30 | 2020-11-30 | Fusion rainfall forecasting method based on multi-model integration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011372595.5A CN113267834A (en) | 2020-11-30 | 2020-11-30 | Fusion rainfall forecasting method based on multi-model integration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113267834A true CN113267834A (en) | 2021-08-17 |
Family
ID=77227843
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011372595.5A Pending CN113267834A (en) | 2020-11-30 | 2020-11-30 | Fusion rainfall forecasting method based on multi-model integration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113267834A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114563834A (en) * | 2022-04-27 | 2022-05-31 | 知一航宇(北京)科技有限公司 | Numerical forecast product interpretation application method and system |
| CN114626577A (en) * | 2022-02-21 | 2022-06-14 | 北京大学 | Method for forecasting precipitation phase state in winter by artificial intelligence |
| CN115236770A (en) * | 2022-06-29 | 2022-10-25 | 广西壮族自治区气象科学研究所 | Nonlinear short-time approach precipitation forecasting method based on space-time stacking and sample reconstruction |
| WO2023162482A1 (en) * | 2022-02-25 | 2023-08-31 | 株式会社天地人 | Precipitation prediction system, precipitation prediction method, program, base station selection system, and base station selection method |
| CN117111181A (en) * | 2023-09-05 | 2023-11-24 | 浙江省气象台 | Short-time strong precipitation probability prediction method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110579823A (en) * | 2019-09-02 | 2019-12-17 | 中国电力科学研究院有限公司 | method and system for forecasting short-term rainfall |
| CN110967695A (en) * | 2019-10-28 | 2020-04-07 | 兰州大方电子有限责任公司 | Radar echo extrapolation short-term prediction method based on deep learning |
-
2020
- 2020-11-30 CN CN202011372595.5A patent/CN113267834A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110579823A (en) * | 2019-09-02 | 2019-12-17 | 中国电力科学研究院有限公司 | method and system for forecasting short-term rainfall |
| CN110967695A (en) * | 2019-10-28 | 2020-04-07 | 兰州大方电子有限责任公司 | Radar echo extrapolation short-term prediction method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 程丛兰,陈明轩,王建捷,高峰,杨汉贤: "基于雷达外推临近预报和中尺度数值预报融合技术的短时定量降水预报试验", 《气象学报》 * |
| 龙清怡,刘海文,顾建峰,张亚萍,翟丹华,杨春: "雷达资料与中尺度数值预报的融合方法研究及其在临近预报中的应用", 《气象》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114626577A (en) * | 2022-02-21 | 2022-06-14 | 北京大学 | Method for forecasting precipitation phase state in winter by artificial intelligence |
| CN114626577B (en) * | 2022-02-21 | 2023-05-12 | 北京大学 | Method for forecasting winter precipitation phase state by utilizing artificial intelligence |
| WO2023162482A1 (en) * | 2022-02-25 | 2023-08-31 | 株式会社天地人 | Precipitation prediction system, precipitation prediction method, program, base station selection system, and base station selection method |
| CN114563834A (en) * | 2022-04-27 | 2022-05-31 | 知一航宇(北京)科技有限公司 | Numerical forecast product interpretation application method and system |
| CN114563834B (en) * | 2022-04-27 | 2022-07-26 | 知一航宇(北京)科技有限公司 | Numerical forecast product interpretation application method and system |
| CN115236770A (en) * | 2022-06-29 | 2022-10-25 | 广西壮族自治区气象科学研究所 | Nonlinear short-time approach precipitation forecasting method based on space-time stacking and sample reconstruction |
| CN115236770B (en) * | 2022-06-29 | 2024-05-28 | 广西壮族自治区气象科学研究所 | Nonlinear short-time adjacent precipitation prediction method based on space-time stacking and sample reconstruction |
| CN117111181A (en) * | 2023-09-05 | 2023-11-24 | 浙江省气象台 | Short-time strong precipitation probability prediction method and system |
| CN117111181B (en) * | 2023-09-05 | 2024-04-09 | 浙江省气象台 | Short-time strong precipitation probability prediction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113267834A (en) | Fusion rainfall forecasting method based on multi-model integration | |
| CN112213727B (en) | Precipitation correction method of satellite-borne radar based on active and passive microwave combined detection | |
| Chumchean et al. | Correcting of real-time radar rainfall bias using a Kalman filtering approach | |
| CN110346844A (en) | Quantitative Precipitation estimating and measuring method of the NRIET based on cloud classification and machine learning | |
| CN111308581B (en) | Radar-rain gauge combined rainfall estimation method based on space-time local model | |
| CN112070286B (en) | Precipitation forecast and early warning system for complex terrain river basin | |
| CN109919353A (en) | A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence | |
| KR102006847B1 (en) | System and Method for radar based nowcasting using optical flow with a multi scale strategy | |
| CN109063083B (en) | Multi-source meteorological information data assimilation method | |
| KR100921424B1 (en) | Method of rainfall estimation using doppler weather radar | |
| CN108535731A (en) | It is short to face precipitation forecast method and device | |
| CN109085664A (en) | A kind of fine temperature forecast deviation sliding correction method | |
| KR101994387B1 (en) | System and Method for Forecasting Precipitation in Mountainous Areas using Water Vapor Weighting Blending | |
| CN111474529A (en) | Method for inverting radar echo by satellite, system for inverting radar echo and navigation radar | |
| CN113281754A (en) | WRF-Hydro key parameter calibration method for quantitatively estimating rainfall by integrating rainfall station with radar | |
| KR101486798B1 (en) | Step-wise scaling method for correcting bias of climate information | |
| Gerstner et al. | Real-time areal precipitation determination from radar by means of statistical objective analysis | |
| CN114898206A (en) | Short-time heavy rainfall forecasting method, computer equipment and storage medium | |
| CN112363140A (en) | Thermodynamic constraint extrapolation objective correction method based on cyclic neural network | |
| KR20150127424A (en) | Method of adjusting radar precipitation estimation, recording medium and device for performing the method | |
| CN116609859A (en) | Weather disaster high-resolution regional mode forecasting system and method | |
| CN115685396A (en) | Rainfall forecast correction method and device | |
| CN109283505A (en) | A method of amendment Radar Echo Extrapolation image Divergent Phenomenon | |
| CN114004426B (en) | Dynamic adjustment method of short-time rainstorm forecast release model | |
| Pejcic et al. | Polarimetric radar-based methods for evaluation of hydrometeor mixtures in numerical weather prediction models |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210817 |
|
| RJ01 | Rejection of invention patent application after publication |