CN116975523A - Data assimilation background error covariance characteristic statistical method for strong convection weather typing - Google Patents
Data assimilation background error covariance characteristic statistical method for strong convection weather typing Download PDFInfo
- Publication number
- CN116975523A CN116975523A CN202311231702.6A CN202311231702A CN116975523A CN 116975523 A CN116975523 A CN 116975523A CN 202311231702 A CN202311231702 A CN 202311231702A CN 116975523 A CN116975523 A CN 116975523A
- Authority
- CN
- China
- Prior art keywords
- field
- convection
- strong convection
- correlation
- background
- 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.)
- Granted
Links
- 238000007619 statistical method Methods 0.000 title claims abstract description 8
- 238000000034 method Methods 0.000 claims abstract description 26
- 238000001914 filtration Methods 0.000 claims abstract description 16
- 238000001228 spectrum Methods 0.000 claims abstract description 10
- 238000011160 research Methods 0.000 claims abstract description 8
- 238000005516 engineering process Methods 0.000 claims abstract description 4
- 238000001556 precipitation Methods 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000013473 artificial intelligence Methods 0.000 claims description 7
- 238000000926 separation method Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000001360 synchronised effect Effects 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000002310 reflectometry Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- 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
Abstract
The invention discloses a data assimilation background error covariance characteristic statistical method for strong convection weather typing, which relates to the field of atmospheric science research and comprises the following steps: step one, parting the strong convection weather situation of a research area; judging whether the new prediction field is a strong convection background field according to the spatial similarity criterion, if so, further judging which type belongs to the type classified in the step one so as to classify the current new prediction field; and thirdly, after classifying the new prediction field, selecting all historical samples under the type, and then separating the global mode large-scale background error covariance from the regional mode small-scale background error covariance of the sample prediction field based on a frequency spectrum filtering technology, so as to analyze the background error covariance multi-scale characteristic differences of a convection region and a non-convection region under different strong convection weather types. The invention improves the description capability of the numerical mode background error by improving the background error covariance assimilation method.
Description
Technical Field
The invention relates to the field of atmospheric science research, in particular to a data assimilation background error covariance characteristic statistical method for strong convection weather typing.
Background
The strong convection weather has the characteristics of rapid development, rapid movement, strong locality and the like, and the high-resolution numerical prediction is an important way for improving the strong convection weather prediction capability, and the prediction performance of the strong convection weather is greatly dependent on the improvement of the initial field of the mode of observation data assimilation (Thiruvengadam et al 2020). The combination of advanced assimilation method to promote strong convection forecast is the development trend of quantitative application of observation data. In recent years, four-dimensional set-variant mixed assimilation techniques have been fully developed in business and scientific institutions such as europe and the united states, and the "flow dependent" background error covariance provides an important aid for effective observation (Dutta et al 2016; honda et al 2018), but has a balance between performance and computational cost. Because of the diversity of the background situation of strong convection weather, the construction of key information of background error covariance needs reasonable error adaptability representation capability for atmospheric elements, cloud water information and ground surface information. In addition, the average background error covariance cannot distinguish the difference of the error characteristics of the convection region and the non-convection region.
How to observe the thermodynamic and dynamic structures of an initial field in an improved mode more effectively and coordinately in the assimilation process at high frequency plays an important role in strong convection weather forecast, and the key of the problem is the description capability of background error covariance on the spatial characteristics of thermodynamic and dynamic variables and the relation characteristics among variables under different strong convection weather background situations. Considering the schemes of three-dimensional variation, four-dimensional variation, collection assimilation, mixing assimilation and the like of the current main stream, the assimilation methods have advantages and limitations in assimilation capability and calculation efficiency, the assimilation methods cannot fully reflect diversified strong convection weather characteristics, and the average background error covariance cannot distinguish the difference of error characteristics of a convection area and a non-convection area, so that how to coordinate high-frequency observation data faces challenges in improving a mode of thermal power and a power initial field.
Therefore, it is necessary to develop a background error covariance construction method adapting to strong convection weather characteristics, and we want to develop a background error covariance model with different strong convection weather type recognition capabilities and capable of distinguishing convection area and non-convection area characteristics based on artificial intelligence algorithm, so as to improve the coordination effect of observation materials on the thermal and dynamic structures of the initial field.
Disclosure of Invention
The invention aims to: in order to solve the technical defects, the invention provides a data assimilation background error covariance characteristic statistical method for strong convection weather typing, which aims to improve the coordination effect of observed data on a thermodynamic and dynamic structure of a mode initial field, thereby improving the description capability of a numerical mode background error.
According to the method, the numerical mode background error covariance multiscale characteristics of the strong convection weather event are analyzed, the advantages of an artificial intelligence algorithm are combined, and the scene distinguishing and application capacity of the static background error covariance based on the climate state is improved. The method can embody the adaptability of weather situation in the covariance of the background error of the assimilation of the atmospheric data, and improves the description capability of the numerical mode background error by improving the coordination effect of the observation data on the thermodynamic and dynamic structures of the mode initial field.
The technical scheme is as follows: a strong convection weather parting data assimilation background error covariance characteristic statistical method comprises the following steps:
step one, parting the strong convection weather situation of a research area;
judging whether the new prediction field is a strong convection background field according to the spatial similarity criterion, if so, further judging which type belongs to the type classified in the step one so as to classify the current new prediction field;
and thirdly, after classifying the new prediction field, selecting all historical samples under the type, and then separating the global mode large-scale background error covariance from the regional mode small-scale background error covariance of the sample prediction field based on a frequency spectrum filtering technology, so as to analyze the background error covariance multi-scale characteristic differences of a convection region and a non-convection region under different strong convection weather types.
Further, the specific operation of the strong convection weather situation typing in the first step is as follows:
step 1.1, dividing geographic areas of a research range according to the types of strong convection weather systems, then respectively sorting convection samples of different geographic areas, and extracting a plurality of physical quantity fields of synchronous ERA5 re-analysis data; the physical quantity field comprises potential heights, latitudinal winds, longitudinal winds, vertical speeds, temperatures and relative humidity on different isobaric surfaces;
step 1.2, sorting strong convection samples of different geographic areas;
and 1.3, selecting proper physical factors in synchronization, preprocessing data of strong convection examples, and typing background potential fields of the strong convection examples in different geographic areas.
Further, in the step 1.3, an artificial intelligence algorithm is adopted when the background potential field is typed, and the algorithm formula is as follows:wherein (1)>For the original data +.>Score as principal component, ->Is the main component load;
the algorithm is used for arranging different grid point values of the data R to be typed according to 'rows', arranging values of the same grid point at different times according to 'columns', and then decomposing the values into two low-dimensional matrixes C and L; the principal component load L is a eigenvector of the data correlation matrix and scaled according to the square root of the corresponding eigenvalue; then, sorting all the main component scores C in descending order according to the sizes of the corresponding characteristic values; and finally, taking a principal component score corresponding to a characteristic value with the accumulated contribution rate exceeding a certain percentage to the original data R, and performing oblique rotation on the principal component score C to obtain a better solution.
Further, in the second step, the spatial similarity judging method is as follows:
step 2.1, respectively calculating the spatial similarity between the latest forecasting field generated by the numerical forecasting mode and the strong convection weather potential fields of various types obtained by the parting in the step one;
step 2.2, if the similarity values of the new forecasting field and all strong convection weather situation fields are smaller than 0.9, judging that the current new forecasting field is not an environmental background field with strong convection weather;
otherwise, the new prediction field is a strong convection background field, and then the next step of judgment is carried out;
step 2.3, if the similarity value between the new forecast field and a certain strong convection weather shape potential field is the highest and is more than or equal to 0.9, classifying the current weather background into the type;
the algorithm formula adopted is as follows:wherein (1)>Is a new forecasting field and the firstjSpatial similarity between weather conditions of class strong convection,/->Is the total lattice count of the spatial field, +.>And->Respectively the firstiNew forecast value level and the th of each grid pointjBackground field distance flat value of class strong convection weather type.
Further, the specific operation steps of the third step are as follows:
step 3.1, distinguishing a convection area and a non-convection area in a history strong convection example forecasting field by taking whether the convection precipitation is a criterion;
step 3.2, respectively aiming at the convection area and the non-convection area, utilizing a frequency spectrum filtering algorithm to separate the spatial scale:wherein (1)>And->Respectively are lattice pointsnAnd its previous new estimate of the grid point,an original value for the lattice point; />As weight parameter and filtering timeskCharacteristic wavelength->Space between gridsdRelated to; during each filtering, the formula (4) needs to be repeatedly applied for two times alternately along the directions of rows and columns so as to ensure the symmetry of analysis; after spectrum filtering, obtaining mode prediction fields with different spatial scales;
and 3.3, judging the logic consistency of the correlation coefficient and the prediction error according to different spatial scales in the convection region and the non-convection region respectively, and determining the horizontal correlation length and the vertical correlation length.
Further, in the step 3.1, the method for dividing the convection zone from the non-convection zone is as follows:
convective precipitation is considered to occur when a signal exceeding 39 dBZ appears in the radar echo;
inversion formula according to radar echo and rainfall intensityWhere mm/hr represents the hour rainfall and dBZ is radar reflectivity, it is known that the precipitation intensity corresponding to 39 dBZ is 9.98 mm/h, so if the precipitation intensity is greater than 9.98 mm/h, it is defined as convective precipitation, and the precipitation area smaller than this intensity is defined as non-convective area.
Further, the specific operation procedure of step 3.3 is as follows:
step 3.3.1, calculating horizontal correlation coefficients between each grid point and all grid points in the field and correlation coefficients between a plurality of sag heights of each grid point in the mode field with scale separation in the convection area and the mode field with scale separation in the non-convection area respectively:
the calculation formula of the horizontal autocorrelation coefficient is as follows:wherein (1)>Horizontal correlation coefficients calculated based on set forecast between A, B grid points of the same height in a mode field, M is total number of set members, +.>And->Forecast values for the ith set member of the A, B two lattice points, respectively, +.>And->Respectively predicting average values for the set of A, B two lattice points;
the vertical autocorrelation coefficient is calculated as follows:wherein (1)>Vertical correlation coefficients calculated for a certain lattice X, Y in the mode field between two different heights based on set predictions, M being the total number of set members, +.>And->Forecast values for the ith set member on the X and Y elevation fields of the lattice, respectively, +.>And->Forecasting average values for the X-th and Y-th height fields of the grid point respectively;
step 3.3.2, comparing the average forecast value of the mode set with the deviation of the observation value of the corresponding scale, and judging whether the correlation is correct or not;
assume that the forecast deviations at any two different grid points of the same height in the mode field are respectivelyThe horizontal correlation coefficient between these two lattice points is +.>The method comprises the steps of carrying out a first treatment on the surface of the When->When the logic consistency criterion is met; that is, a logical relationship is established only if the predictions for two points are both less than or greater than the observation and there is a positive correlation between the two points, or if the prediction for one point is greater than the observation and the prediction for the other point is less than the observation and there is a negative correlation between the two points; the vertical correlation is the same as the prediction error logical consistency;
step 3.3.3, calculating the proportion of samples with established logic relations in the statistical samples, and carrying out significance test on the correlation coefficient; after the correlation passes the significance test, each lattice point is taken as a central point, and surrounding lattice points with the reliability of the correlation of more than 80% are searched, so that the horizontal and vertical correlation length scales of the central point are determined;
the significance test formula is as follows:wherein (1)>For the correlation coefficient +.>Is the total lattice count of the spatial field; first determine the significance level +.>And find the corresponding threshold +.>Then, an assumption is made that there is no correlation between the two lattice points; calculate statistics +.>If->The original assumption is rejected, i.e. there is a correlation between the two lattice points and vice versa.
The beneficial effects are that: the invention improves the coordination effect of the observed data on the thermodynamic and dynamic structures of the mode initial field, and improves the description capability of the numerical mode background error by improving the background error covariance assimilation method.
Drawings
FIG. 1 is a flow chart of a method for data assimilation background error covariance feature statistics for strong convection weather typing;
FIG. 2 is a graph of the strongly convective weather pattern typing results in the east China;
FIG. 3 is a graph of aggregate background error covariance multi-scale feature distribution at different spatial scales.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
The invention provides a data assimilation background error covariance characteristic statistical method for strong convection weather typing, which is shown in a flow chart as shown in figure 1 and specifically comprises the following steps:
step 1.1, geographical area division:
the eastern region of China is a strong convection region, and different geographic areas are affected by different strong convection weather systems. The strong convection weather system mainly comprises a shear line, a subtropical high pressure, a low-altitude rapid flow, a high-altitude groove, a cold vortex, a low vortex shear and the like.
Therefore, the eastern region of China is divided into three regions of east China, south China and north China according to geographic demarcations; specifically, the research ranges in this embodiment are 104-123°e,18-44°n, and they are divided into three areas of eastern China, south China and north China, each of which ranges are as follows: the range of the east China is 114-123 DEG E,26-39 DEG N; the range of the south China is 104-118 DEG E,18-27 DEG N; the North China area is 110-120 DEG E and 34-44 DEG N. Then, respectively sorting the convection samples in each region, and extracting a plurality of physical quantity fields of the synchronous ERA5 re-analysis data; the physical quantity fields include potential heights on different isobaric surfaces, weft direction wind, warp direction wind, vertical velocity, temperature and relative humidity.
In this embodiment, convection samples in 2012-2022 of three regions including east China, south China and north China are selected, and a plurality of physical quantity fields of synchronous ERA5 re-analysis data are extracted, wherein the physical quantity fields mainly comprise potential height, latitudinal wind, longitudinal wind, vertical speed, temperature and relative humidity on equal pressure surfaces of 850hPa, 700hPa and 500 hPa.
Step 1.2, extracting strong convection examples;
extracting strong convection samples of different areas of east China, south China and north China;
step 1.3, screening physical factors to be typed and preprocessing data;
selecting proper physical factors in the same period, such as potential height, weft direction wind, warp direction wind, vertical speed, temperature and relative humidity on a plurality of isobaric surfaces, and carrying out data preprocessing on the physical factors, namely, arranging different grid point values of the same physical factors in rows, arranging different time values of the same grid point in columns, and then carrying out background shape potential field typing on strong convection examples in different areas of east China, south China and north China;
and when the background shape potential field typing is carried out, an artificial intelligence algorithm is adopted, and the weather background situation corresponding to the history strong convection sample is divided into a plurality of different types by using the artificial intelligence algorithm.
The artificial intelligence algorithm is explained specifically as follows:wherein (1)>For the original data +.>Score as principal component, ->Is the principal component load.
The method is to sort data to be typedThe different lattice values of the same lattice are arranged in "rows", the values of the same lattice at different times are arranged in "columns", and then they are decomposed into +.>And->Two low-dimensional matrices. Principal component load->Is a eigenvector of the data correlation matrix and scaled by the square root of its corresponding eigenvalue. Then scoring all principal components according to the size of the corresponding characteristic value +.>Sorting in descending order, and finally taking the original data +.>Adding up the principal component scores corresponding to the characteristic values with the contribution rate exceeding a certain percentage, and scoring the principal component +.>A tilt rotation is performed to obtain a more optimal solution.
As shown in fig. 2, there are 9 different types of strong convection weather profiles in the 850hPa eastern region (white box). The arrow indicates the wind direction, the solid black line indicates the temperature field, and the upper left hand corner number indicates the interpretation variance and number of samples, respectively, for that weather type. The 9 different types of strong convection weather conditions are: the first type indicates that the east China is subjected to cold and low pressure control, southwest wind is strong, and warm and humid air flow is conveyed to the east China; the second type shows that the east China is controlled by southwest air flow at the northwest side of the auxiliary high, and the low-layer temperature is higher; the third type represents the intersection of cold air from north and warm air from south in eastern China; the fourth type represents the heated low pressure control in the eastern China; the fifth type represents northwest cold air flow control after being slotted in eastern China; the sixth type represents the most prevailing southern wind in the eastern China, for which water vapor is transported; the seventh type indicates that the eastern China is affected by the shear line, and southwest wind is converted into southeast wind; the eighth type represents east side airflow control in the eastern region of China under cold, low pressure; the ninth type indicates that eastern China is poorly controlled by northeast air flow.
Judging whether the new prediction field is a strong convection background field according to the spatial similarity criterion, if so, further judging which type belongs to the first step of parting so as to classify the current new prediction field;
step 2.1, respectively calculating the spatial similarity between the latest forecasting field generated by the numerical forecasting mode and the strong convection weather potential fields of various types obtained by the parting in the step one;
step 2.2, if the similarity values of the new forecasting field and all strong convection weather situation fields are smaller than 0.9, judging that the current new forecasting field is not an environmental background field with strong convection weather;
otherwise, the new prediction field is a strong convection background field, and then the next step of judgment is carried out;
and 2.3, if the similarity value between the new forecast field and a certain strong convection weather shape potential field is the highest and is more than or equal to 0.9, the current weather background is classified as the type.
The specific algorithm is explained as follows:wherein (1)>Is a new forecasting field and the firstjSpatial similarity between weather conditions of class strong convection,/->Is the total lattice count of the spatial field, +.>And->Respectively the firstiNew forecast value level and the th of each grid pointjBackground field distance flat value of class strong convection weather type.
In fig. 1, after a new prediction field is classified, all historical samples under the type are selected, and then a global mode large-scale background error covariance and a regional mode small-scale background error covariance are separated from a sample prediction field based on a frequency spectrum filtering technology, so that background error covariance multi-scale characteristic differences of a convection region and a non-convection region under different strong convection weather types are analyzed.
The specific operation steps of the third step are as follows:
step 3.1, distinguishing a convection area and a non-convection area in a history strong convection example forecasting field by taking whether the convection precipitation is a criterion or not:
convective precipitation is generally considered to occur when signals exceeding 39 dBZ appear in the radar echo. Inversion formula according to radar echo and rainfall intensityWherein mm/hr represents an hour rainfalldBZ is radar reflectivity, and the precipitation intensity corresponding to 39 dBZ is 9.98 mm/h, so if the precipitation intensity is more than 9.98 mm/h, the precipitation is defined as convection precipitation, and the precipitation area smaller than the intensity is defined as a non-convection area.
Step 3.2, respectively aiming at the convection area and the non-convection area, utilizing a frequency spectrum filtering algorithm to separate the spatial scale:wherein (1)>And->Respectively are lattice pointsnAnd its previous new estimate of the grid point,an original value for the lattice point; />As weight parameter and filtering timeskCharacteristic wavelength->Space between gridsdRelated to the following. Each time filtering, equation (4) needs to be applied alternately back and forth in the row and column directions twice to ensure symmetry of the analysis. After spectrum filtering, different spatial scales such as large scale is obtained>1000 km), mesoscale (100-1000 km).
And 3.3, judging the logic consistency of the correlation coefficient and the prediction error according to different spatial scales in the convection region and the non-convection region respectively, and determining the horizontal correlation length and the vertical correlation length.
Step 3.3.1, calculating horizontal correlation coefficients between each grid point and all grid points in the field and correlation coefficients between a plurality of sag heights of each grid point in the mode field with scale separation in the convection area and the mode field with scale separation in the non-convection area respectively:
horizontal autocorrelation coefficient meterThe calculation formula is as follows:wherein (1)>Horizontal correlation coefficients calculated based on set forecast between A, B grid points of the same height in a mode field, M is total number of set members, +.>And->Forecast values for the ith set member of the A, B two lattice points, respectively, +.>And->The average value is predicted for each of the sets of A, B grid points.
The vertical autocorrelation coefficient is calculated as follows:wherein (1)>Vertical correlation coefficients calculated for a certain lattice X, Y in the mode field between two different heights based on set predictions, M being the total number of set members, +.>And->Forecast values for the ith set member on the X and Y elevation fields of the lattice, respectively, +.>And->The X and Y heights of the lattice point respectivelyThe set of degree fields predicts the average.
Step 3.3.2, comparing the average forecast value of the mode set with the deviation of the observation value of the corresponding scale, and judging whether the correlation is correct or not;
assume that the forecast deviations at any two different grid points of the same height in the mode field are respectivelyThe horizontal correlation coefficient between these two lattice points is +.>The method comprises the steps of carrying out a first treatment on the surface of the When->When the logic consistency criterion is met; that is, a logical relationship is established only if the predictions for two points are both less than or greater than the observation and there is a positive correlation between the two points, or if the prediction for one point is greater than the observation and the prediction for the other point is less than the observation and there is a negative correlation between the two points; the vertical correlation is the same as the prediction error logical consistency;
step 3.3.3, calculating the proportion of samples with established logic relations in the statistical samples, and carrying out significance test on the correlation coefficient; after the correlation passes the significance test, each lattice point is taken as a central point, and surrounding lattice points with the reliability of the correlation of more than 80% are searched, so that the horizontal and vertical correlation length scales of the central point are determined;
the significance test formula is as follows:wherein (1)>For the correlation coefficient +.>Is the total lattice count of the spatial field; first determine the significance level +.>And find out the corresponding clinicBoundary value->Then, an assumption is made that there is no correlation between the two lattice points; calculate statistics +.>If->The original assumption is rejected, i.e. there is a correlation between the two lattice points and vice versa.
As shown in fig. 3, taking the tokyo (119°e,32°n) as an example, the large scale (a-c) and medium scale (d-f) background error covariance horizontal feature profiles, the range formed by the 0.7 contour is the horizontal length scale of the point in each direction. After the autocorrelation coefficient passes the 0.05 significance test, the situation that the correlation of many remote false signals is low is eliminated. Mesoscale horizontal features are similar to large scale horizontal features, but are much smaller in scope, exhibiting localized features. Furthermore, as the height increases, the relevant length increases slightly, but this feature is not apparent on a large scale.
As described above, although the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limiting the invention itself. Various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. A strong convection weather parting data assimilation background error covariance characteristic statistical method is characterized by comprising the following steps:
step one, parting the strong convection weather situation of a research area;
judging whether the new prediction field is a strong convection background field according to the spatial similarity criterion, if so, further judging which type belongs to the type classified in the step one so as to classify the current new prediction field;
and thirdly, after classifying the new prediction field, selecting all historical samples under the type, and then separating the global mode large-scale background error covariance from the regional mode small-scale background error covariance of the sample prediction field based on a frequency spectrum filtering technology, so as to analyze the background error covariance multi-scale characteristic differences of a convection region and a non-convection region under different strong convection weather types.
2. The method for counting the covariance characteristics of the data assimilation background errors of the strong convection weather typing according to claim 1, wherein the specific operation of the strong convection weather typing in the step one is as follows:
step 1.1, dividing geographic areas of a research range according to the types of strong convection weather systems, then respectively sorting convection samples of different geographic areas, and extracting a plurality of physical quantity fields of synchronous ERA5 re-analysis data; the physical quantity field comprises potential heights, latitudinal winds, longitudinal winds, vertical speeds, temperatures and relative humidity on different isobaric surfaces;
step 1.2, sorting strong convection samples of different geographical areas;
and 1.3, selecting proper physical factors in synchronization, preprocessing data of strong convection examples, and typing background potential fields of the strong convection examples in different geographic areas.
3. The method for counting the data assimilation background error covariance features of strong convection weather typing according to claim 2, wherein in the step 1.3, an artificial intelligence algorithm is adopted when the background potential field typing is carried out, and the algorithm formula is as follows:wherein (1)>For the original data +.>Score as principal component, ->Is the main component load;
the algorithm is used for arranging different grid point values of the data R to be typed according to 'rows', arranging values of the same grid point at different times according to 'columns', and then decomposing the values into two low-dimensional matrixes C and L; the principal component load L is a eigenvector of the data correlation matrix and scaled according to the square root of the corresponding eigenvalue; then, sorting all the main component scores C in descending order according to the sizes of the corresponding characteristic values; and finally, taking a principal component score corresponding to a characteristic value with the accumulated contribution rate exceeding a certain percentage to the original data R, and performing oblique rotation on the principal component score C to obtain a better solution.
4. The method for counting the covariance characteristics of the data assimilation background error of strong convection weather typing according to claim 1, wherein in the second step, the spatial similarity judging method is as follows:
step 2.1, respectively calculating the spatial similarity between the latest forecasting field generated by the numerical forecasting mode and the strong convection weather potential fields of various types obtained by the parting in the step one;
step 2.2, if the similarity values of the new forecasting field and all strong convection weather situation fields are smaller than 0.9, judging that the current new forecasting field is not an environmental background field with strong convection weather;
otherwise, the new prediction field is a strong convection background field, and then the next step of judgment is carried out;
step 2.3, if the similarity value between the new forecast field and a certain strong convection weather shape potential field is the highest and is more than or equal to 0.9, classifying the current weather background into the type;
the algorithm formula adopted is as follows:wherein (1)>Is a new forecasting field and the firstjSpatial similarity between class strong convection weather conditions,/>is the total lattice count of the spatial field, +.>And->Respectively the firstiNew forecast value level and the th of each grid pointjBackground field distance flat value of class strong convection weather type.
5. The method for counting the covariance characteristics of the data assimilation background error of strong convection weather typing according to claim 1, wherein the specific operation steps of the third step are as follows:
step 3.1, distinguishing a convection area and a non-convection area in a history strong convection example forecasting field by taking whether the convection precipitation is a criterion;
step 3.2, respectively aiming at the convection area and the non-convection area, utilizing a frequency spectrum filtering algorithm to separate the spatial scale:wherein (1)>And->Respectively are lattice pointsnAnd its previous new estimate, +.>An original value for the lattice point; />As weight parameter and filtering timeskCharacteristic wavelength->Space between gridsdRelated to; during each filtering, the formula (4) needs to be repeatedly applied for two times alternately along the directions of rows and columns so as to ensure the symmetry of analysis; after spectrum filtering, obtaining mode prediction fields with different spatial scales;
and 3.3, judging the logic consistency of the correlation coefficient and the prediction error according to different spatial scales in the convection region and the non-convection region respectively, and determining the horizontal correlation length and the vertical correlation length.
6. The method for counting the covariance characteristics of the data assimilation background error of strong convection weather typing according to claim 5, wherein in step 3.1, the method for dividing convection regions from non-convection regions is as follows:
convective precipitation is considered to occur when a signal exceeding 39 dBZ appears in the radar echo;
inversion formula according to radar echo and rainfall intensityWhere mm/hr represents the hour rainfall and dBZ is radar reflectivity, it is known that the precipitation intensity corresponding to 39 dBZ is 9.98 mm/h, so if the precipitation intensity is greater than 9.98 mm/h, it is defined as convective precipitation, and the precipitation area smaller than this intensity is defined as non-convective area.
7. The method for counting the covariance characteristics of the data assimilation background error for strong convection weather typing according to claim 5, wherein the specific operation procedure of the step 3.3 is as follows:
step 3.3.1, calculating horizontal correlation coefficients between each grid point and all grid points in the field and correlation coefficients between a plurality of sag heights of each grid point in the mode field with scale separation in the convection area and the mode field with scale separation in the non-convection area respectively:
the calculation formula of the horizontal autocorrelation coefficient is as follows:wherein (1)>Horizontal correlation coefficients calculated based on set forecast between A, B grid points of the same height in a mode field, M is total number of set members, +.>And->Forecast values for the ith set member of the A, B two lattice points, respectively, +.>And->Respectively predicting average values for the set of A, B two lattice points;
the vertical autocorrelation coefficient is calculated as follows:wherein (1)>Vertical correlation coefficients calculated for a certain lattice X, Y in the mode field between two different heights based on set predictions, M being the total number of set members, +.>And->Forecast values for the ith set member on the X and Y elevation fields of the lattice, respectively, +.>And->Forecasting average values for the X-th and Y-th height fields of the grid point respectively;
step 3.3.2, comparing the average forecast value of the mode set with the deviation of the observation value of the corresponding scale, and judging whether the correlation is correct or not;
assume that the forecast deviations at any two different grid points of the same height in the mode field are respectivelyThe horizontal correlation coefficient between these two lattice points is +.>The method comprises the steps of carrying out a first treatment on the surface of the When->When the logic consistency criterion is met; that is, a logical relationship is established only if the predictions for two points are both less than or greater than the observation and there is a positive correlation between the two points, or if the prediction for one point is greater than the observation and the prediction for the other point is less than the observation and there is a negative correlation between the two points; the vertical correlation is the same as the prediction error logical consistency;
step 3.3.3, calculating the proportion of samples with established logic relations in the statistical samples, and carrying out significance test on the correlation coefficient; after the correlation passes the significance test, each lattice point is taken as a central point, and surrounding lattice points with the reliability of the correlation of more than 80% are searched, so that the horizontal and vertical correlation length scales of the central point are determined;
the significance test formula is as follows:wherein (1)>For the correlation coefficient +.>Is the total lattice count of the spatial field; first determine the significance level +.>And find the corresponding threshold +.>Then, an assumption is made that there is no correlation between the two lattice points; calculate statistics +.>If->The original assumption is rejected, i.e. there is a correlation between the two lattice points and vice versa.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311231702.6A CN116975523B (en) | 2023-09-22 | 2023-09-22 | Data assimilation background error covariance characteristic statistical method for strong convection weather typing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311231702.6A CN116975523B (en) | 2023-09-22 | 2023-09-22 | Data assimilation background error covariance characteristic statistical method for strong convection weather typing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116975523A true CN116975523A (en) | 2023-10-31 |
| CN116975523B CN116975523B (en) | 2023-12-12 |
Family
ID=88473395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311231702.6A Active CN116975523B (en) | 2023-09-22 | 2023-09-22 | Data assimilation background error covariance characteristic statistical method for strong convection weather typing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116975523B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992057A (en) * | 2015-06-25 | 2015-10-21 | 南京信息工程大学 | Quasi-ensemble-variation based mixed data assimilation method |
| CN110110922A (en) * | 2019-04-30 | 2019-08-09 | 南京信息工程大学 | A kind of adaptive partition assimilation method based on rain belt sorting technique |
| CN110968926A (en) * | 2018-09-29 | 2020-04-07 | 中国科学院国家空间科学中心 | Method for predicting atmospheric parameters based on improved background error covariance matrix |
| CN113740220A (en) * | 2021-09-07 | 2021-12-03 | 中国人民解放军国防科技大学 | Multi-scale three-dimensional variational assimilation method based on high-resolution aerosol data |
| CN113792262A (en) * | 2021-09-27 | 2021-12-14 | 王世璋 | Method for assimilating local collection data |
| US20230048788A1 (en) * | 2021-08-06 | 2023-02-16 | The Government Of The United States Of America, As Represented By The Secretary Of The Navy | Method and System for Multi-scale Assimilation of Surface Water Ocean Topography (SWOT) Observations |
| CN115857056A (en) * | 2022-11-18 | 2023-03-28 | 广东电网有限责任公司广州供电局 | Weather forecasting method and system for fast updating cyclic assimilation by strong convection |
-
2023
- 2023-09-22 CN CN202311231702.6A patent/CN116975523B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992057A (en) * | 2015-06-25 | 2015-10-21 | 南京信息工程大学 | Quasi-ensemble-variation based mixed data assimilation method |
| CN110968926A (en) * | 2018-09-29 | 2020-04-07 | 中国科学院国家空间科学中心 | Method for predicting atmospheric parameters based on improved background error covariance matrix |
| CN110110922A (en) * | 2019-04-30 | 2019-08-09 | 南京信息工程大学 | A kind of adaptive partition assimilation method based on rain belt sorting technique |
| US20230048788A1 (en) * | 2021-08-06 | 2023-02-16 | The Government Of The United States Of America, As Represented By The Secretary Of The Navy | Method and System for Multi-scale Assimilation of Surface Water Ocean Topography (SWOT) Observations |
| CN113740220A (en) * | 2021-09-07 | 2021-12-03 | 中国人民解放军国防科技大学 | Multi-scale three-dimensional variational assimilation method based on high-resolution aerosol data |
| CN113792262A (en) * | 2021-09-27 | 2021-12-14 | 王世璋 | Method for assimilating local collection data |
| CN115857056A (en) * | 2022-11-18 | 2023-03-28 | 广东电网有限责任公司广州供电局 | Weather forecasting method and system for fast updating cyclic assimilation by strong convection |
Non-Patent Citations (4)
| Title |
|---|
| SIBO CHENG等: "Machine learning with data assimilation and uncertainty quantification for dynamical systems:a review", pages 1 - 26, Retrieved from the Internet <URL:https://arxiv.org/abs/2303.10462> * |
| WANSHUWU等: "Three-Dimensional Variational Analysis with Spatially Inhomogeneous Covariances", 《WUETAL》, pages 2905 - 2916 * |
| 曾燕飞等: "基于随机物理过程扰动的BMJ积云参数化方案对降水集合预报的影响", 《大气科学学报》, vol. 45, no. 3, pages 428 - 443 * |
| 胡婧婷等: "雷达资料不同同化试验对短时降水预报影响的分析", 《暴雨灾害》, vol. 41, no. 6, pages 679 - 690 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116975523B (en) | 2023-12-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lagerquist et al. | Machine learning for real-time prediction of damaging straight-line convective wind | |
| Boettcher et al. | A 10-yr climatology of diabatic Rossby waves in the Northern Hemisphere | |
| CN112800634A (en) | Rainfall estimation method and system coupling dry-wet state identification and multi-source information fusion | |
| CN107133686A (en) | City-level PM2.5 concentration prediction methods based on Spatio-Temporal Data Model for Spatial | |
| Burlando et al. | Mesoscale wind climate analysis: identification of anemological regions and wind regimes | |
| Squitieri et al. | On the forecast sensitivity of MCS cold pools and related features to horizontal grid spacing in convection-allowing WRF simulations | |
| CN106886564A (en) | A kind of method and device that NWP wind energy collection of illustrative plates is corrected based on space clustering | |
| Yang et al. | Quantitative analysis of water vapor transport during Mei-yu front rainstorm period over the Tibetan Plateau and Yangtze-Huai River Basin | |
| Lin et al. | Vertical structure of Tibetan Plateau Vortex in boreal summer | |
| CN114548595A (en) | Strong convection weather physical characteristic quantity prediction method and system based on attention mechanism | |
| Ya-li et al. | Synoptic characteristics related to warm-sector torrential rainfall events in South China during the annually first rainy season | |
| CN114882373A (en) | Multi-feature fusion sandstorm prediction method based on deep neural network | |
| CN116975523B (en) | Data assimilation background error covariance characteristic statistical method for strong convection weather typing | |
| Zhe et al. | A STUDY OF THE INFLUENCE OF MICROPHYSICAL PROCESSES ON TYPHOON NIDA (2016) USING A NEW DOUBLE; MOMENT MICROPHYSICS SCHEME IN THE WEATHER RESEARCH AND FORECASTING MODEL | |
| CN109583095B (en) | North Pacific typhoon extension period forecasting method based on hybrid statistical power model | |
| Wang et al. | Forewarning method of downburst based on feature recognition and extrapolation | |
| CN115511192B (en) | Precipitation prediction method and system based on lightning data assimilation | |
| Steinkohl et al. | Extreme value analysis of multivariate high-frequency wind speed data | |
| Hou et al. | A new method of characterizing flow patterns of vortices and detecting the centers of vortices in a numerical wind field | |
| CN114091741A (en) | Typhoon forecast prediction method and device | |
| Roy et al. | Tropical cyclone and track forecasting | |
| CN113065700A (en) | Short-time heavy rainfall forecasting method based on significance and sensitivity factor analysis method | |
| Erfani et al. | Automated synoptic typing of freezing rain events for hazard analysis | |
| Spassiani | Climatology of severe convective wind gusts in Australia | |
| CN117475297A (en) | Plateau low vortex identification method |
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 |