CN110967695A - Radar echo extrapolation short-term prediction method based on deep learning - Google Patents
Radar echo extrapolation short-term prediction method based on deep learning Download PDFInfo
- Publication number
- CN110967695A CN110967695A CN201911029867.9A CN201911029867A CN110967695A CN 110967695 A CN110967695 A CN 110967695A CN 201911029867 A CN201911029867 A CN 201911029867A CN 110967695 A CN110967695 A CN 110967695A
- Authority
- CN
- China
- Prior art keywords
- radar echo
- radar
- echo data
- memory module
- time
- 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
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000013213 extrapolation Methods 0.000 title claims abstract description 33
- 238000013135 deep learning Methods 0.000 title claims abstract description 8
- 238000002310 reflectometry Methods 0.000 claims abstract description 85
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 5
- 238000001556 precipitation Methods 0.000 claims description 55
- 230000015654 memory Effects 0.000 claims description 54
- 238000002592 echocardiography Methods 0.000 claims description 15
- 238000012935 Averaging Methods 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 230000003247 decreasing effect Effects 0.000 claims description 7
- 238000003908 quality control method Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 230000002146 bilateral effect Effects 0.000 claims description 5
- 230000036961 partial effect Effects 0.000 claims description 5
- 238000013277 forecasting method Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 description 17
- 238000004364 calculation method Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 3
- 230000032683 aging Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000002441 reversible effect Effects 0.000 description 3
- 238000000540 analysis of variance Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000000670 limiting effect Effects 0.000 description 2
- 230000010349 pulsation Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000006886 spatial memory Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/35—Details of non-pulse systems
- G01S7/352—Receivers
- G01S7/354—Extracting wanted echo-signals
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/40—Means for monitoring or calibrating
- G01S7/4052—Means for monitoring or calibrating by simulation of echoes
-
- 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 radar echo extrapolation short-term forecasting method based on deep learning, which belongs to the technical field of climate monitoring and comprises the steps of firstly, obtaining actually measured radar echo data and rainfall; obtaining an equivalent reflectivity factor by using the radar echo data, and obtaining a Z-R relation by using the reflectivity factor and rainfall; constructing a radar echo extrapolation model, preprocessing the radar echo data, training the radar echo extrapolation model by using the processed data, and predicting the radar echo data in a future period of time; finally, inputting the predicted radar echo data into a Z-R relation to obtain the predicted rainfall; the method can forecast the rainfall in a large range in a lattice mode, the rainfall forecast within 1 hour is higher in accuracy than the rainfall forecast in a numerical mode, and the forecast rainfall falling area is more accurate.
Description
Technical Field
The invention relates to the technical field of climate monitoring, in particular to a radar echo extrapolation short-term forecasting method based on deep learning.
Background
Currently, data collected by a radar can be used for judging and predicting weather, and a Doppler radar generally comprises a combined reflectivity factor CR, an echo top height ET, a vertical liquid water content VIL and a 1-hour accumulated precipitation OHP. Each individual sweep reads the values of four types of products on each automated station and regional automated station in real time: and automatically extracting the live precipitation of 1 hour of the large monitoring stations and the automatic stations in the areas of each province, and defining the values of the live precipitation and the live precipitation factor. According to the regulations of different places, the rainfall can be divided into different levels, and the short-time strong rainfall in the Hexi area and the Hedong area is three levels according to the regulations of the short-time strong rainfall local standard of Gansu province: the rainfall is more than or equal to 10mm and less than 20mm in 1 hour; short-time strong precipitation two stages: the rainfall is more than or equal to 20mm and less than 30mm in 1 hour; short-time strong precipitation by one stage: the rainfall is more than or equal to 30mm in 1 hour. Short-time strong precipitation in the east of Hei three levels: the rainfall is more than or equal to 20mm and less than 30mm in 1 hour; short-time strong precipitation two stages: the rainfall is more than or equal to 30mm and less than 50mm in 1 hour; short-time strong precipitation by one stage: the rainfall is more than or equal to 50mm in 1 hour; when rainfall is about to occur or occurs, the rainfall needs to be predicted so as to achieve the purpose of early warning, and when the rainfall is quantitatively estimated by using the default Z-R relation of a radar, an underestimation phenomenon usually exists.
Disclosure of Invention
The invention aims to: the invention provides a radar echo extrapolation short-term forecasting method based on deep learning, which solves the technical problem that rainfall short-term forecasting or forecasting inaccuracy cannot be carried out by utilizing radar data at present.
The technical scheme adopted by the invention is as follows:
a radar echo extrapolation short-term prediction method based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
step 1: acquiring actually measured radar echo data and rainfall;
step 2: obtaining an equivalent reflectivity factor by using the radar echo data, and obtaining a Z-R relation by using the reflectivity factor and rainfall;
and step 3: constructing a radar echo extrapolation model, preprocessing the radar echo data, training the radar echo extrapolation model by using the processed data, and predicting the radar echo data in a future period of time;
and 4, step 4: and inputting the predicted radar echo data into a Z-R relation to obtain the predicted rainfall.
Further, in the step 2, the step of obtaining the reflectivity factor specifically includes:
step 21: processing the radar echo data, including analysis, quality control and effective elevation selection;
step 22: analyzing an equivalent reflectivity factor from the processed radar echo;
step 23: time weighted averaging is performed on the equivalent reflectance factor per x minutes using the formula:
wherein N represents the number of observation data of the radar within 1 hour, N represents the total number of observation data of the radar within 1 hour, and t1,t2,...,tNRespectively representing the time at which the radar measures the point, Z1,Z2,...,ZNRespectively measuring the 1h average equivalent reflectivity factor value of the point by radar;
step 24: converting the equivalent reflectivity factor after the time weighted average into a reflectivity factor, wherein the adopted formula is as follows:
Ze＝10*lgZ (2)，
where Z represents the reflectivity factor and Ze represents the equivalent reflectivity factor after time-weighted averaging.
Further, in the step 2, the specific step of obtaining the Z-R relationship is:
step 25: obtaining a Z-R relation by utilizing the reflectivity factor and the precipitation, wherein the Z-R relation is specifically as follows:
Z＝ARb(3)
wherein Z represents a reflectivity factor, A and b are both parameters, and R represents precipitation;
step 26: and fitting the Z-R relation by using an SPSS tool to obtain the final Z-R relation.
Further, in the step 3, preprocessing the radar echo data includes the following steps
Step 31: inputting the radar echo data into a singular point filter and a bilateral filter:
step 32: inputting the radar echo data obtained in the step 31 into an equivalent reflectivity vertical decreasing rate filter, and filtering ground object echoes and partial super-refraction echoes;
step 33: calculating annual average equivalent reflectivity factor distribution by using the radar echo data obtained in the step 32, determining the position shielded by the radar azimuth angle, and filling the related azimuth angle by linear interpolation;
step 34: constructing a high-pass filter to filter the radar echo data obtained in the step 33 to obtain precipitation echoes;
step 35: and for the precipitation echo, using the equivalent reflectivity of the 0-y layer as the processed radar echo data.
Further, in the step 3, the radar echo extrapolation model adopts a network structure of ST-LSTM-RNN.
Furthermore, the ST-LSTM-RNN network structure comprises an ST-LSTM unit, a time memory module and a space memory module;
inputting the preprocessed radar echo data into a plurality of ST-LSTM units which are connected in sequence, and outputting predicted radar echo data;
the time memory module and the space memory module are added into each ST-LSTM unit, the time memory module is used for memorizing and accumulating the first m moments of the neurons in the current layer, and the space memory module is used for memorizing and superposing the first m moments of the neurons in different layers.
Furthermore, the time memory module and the space memory module respectively comprise an input gate, an output gate and a forgetting gate;
an input gate for deciding what the input at time t needs to be memorized;
the output gate is used for determining the content of the hidden layer output of the ST-LSTM unit;
the forgetting door is used for controlling the memory content to be forgotten;
and fusing the contents of the time memory module and the space memory module to obtain the final hidden layer output content of the ST-LSTM unit.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
the method can forecast the rainfall in a large range in a lattice mode, the rainfall forecast within 1 hour is higher in accuracy than the rainfall forecast in a numerical mode, and the forecast rainfall falling area is more accurate.
The 1h average reflectivity is calculated by a time weight averaging method, and is synthesized by the time weights of the reflectivities observed by all the radars in the hour.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a Z-R fitting relationship chart in example 1 of the present invention;
FIG. 3 is a graph showing the comparison between actual precipitation and reverse precipitation at the Kangnan forest farm in example 1 of the present invention;
FIG. 4 is a comparison between actual precipitation and reverse precipitation in a Linxia Baichuan village in example 1 of the present invention;
FIG. 5 is a diagram illustrating the ST-LSTM-RNN network structure in embodiment 2 of the present invention;
FIG. 6 is a schematic diagram of the structure of the ST-LSTM unit in embodiment 2 of the present invention;
FIG. 7 is a flowchart illustrating radar data preprocessing according to embodiment 2 of the present invention;
FIG. 8 is a graph showing the result of radar data preprocessing (100 KM/range ring) in example 2 of the present invention;
FIG. 9 is the variation rule of skill scores of three sites with forecast time according to the embodiment 2 of the present invention;
FIG. 10 is a comparative plot of the live and forecast of the stormwater process of Lanzhou, 8.8.2018 and 2.7.7 in example 3 of the present invention;
FIG. 11 is a comparison graph of actual precipitation and back-calculated precipitation in the coverage area of the Qingyang radar station in accordance with example 3 of the present invention;
FIG. 12 is a high-level diagram of the coverage area of the Long nan radar in example 4 of the present invention;
fig. 13 is a diagram illustrating elevation selection for the coverage area of the longnan radar in accordance with embodiment 4 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the detailed description and specific examples, while indicating the preferred embodiment of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
It is noted that relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The features and properties of the present invention are described in further detail below with reference to examples.
A radar echo extrapolation short-term prediction method based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
step 1: acquiring actually measured radar echo data and rainfall;
step 2: obtaining an equivalent reflectivity factor by using the radar echo data, and obtaining a Z-R relation by using the reflectivity factor and rainfall;
and step 3: constructing a radar echo extrapolation model, preprocessing the radar echo data, training the radar echo extrapolation model by using the processed data, and predicting the radar echo data in a future period of time;
and 4, step 4: and inputting the predicted radar echo data into a Z-R relation to obtain the predicted rainfall.
In the step 2, the step of obtaining the reflectivity factor specifically includes:
step 21: processing the radar echo data, including analysis, quality control and effective elevation selection;
step 22: analyzing an equivalent reflectivity factor from the processed radar echo;
step 23: time weighted averaging is performed on the equivalent reflectance factor per x minutes using the formula:
wherein N represents the number of observation data of the radar within 1 hour, N represents the total number of observation data of the radar within 1 hour, and t1,t2,...,tNRespectively representing the time at which the radar measures the point, Z1,Z2,...,ZNRespectively measuring the 1h average equivalent reflectivity factor value of the point by radar;
step 24: converting the equivalent reflectivity factor after the time weighted average into a reflectivity factor, wherein the adopted formula is as follows:
Ze＝10*lgZ (5)，
where Z represents the reflectivity factor and Ze represents the equivalent reflectivity factor after time-weighted averaging.
In the step 2, the specific steps for obtaining the Z-R relationship are as follows:
step 25: obtaining a Z-R relation by utilizing the reflectivity factor and the precipitation, wherein the Z-R relation is specifically as follows:
Z＝ARb(6)，
wherein Z represents a reflectivity factor, A and b are both parameters, and R represents precipitation;
step 26: and fitting the Z-R relation by using an SPSS tool to obtain the final Z-R relation.
In the step 3, the preprocessing of the radar echo data comprises the following steps
Step 31: inputting the radar echo data into a singular point filter and a bilateral filter:
step 32: inputting the radar echo data obtained in the step 31 into an equivalent reflectivity vertical decreasing rate filter, and filtering ground object echoes and partial super-refraction echoes;
step 33: calculating annual average equivalent reflectivity factor distribution by using the radar echo data obtained in the step 32, determining the position shielded by the radar azimuth angle, and filling the related azimuth angle by linear interpolation;
step 34: constructing a high-pass filter to filter the radar echo data obtained in the step 33 to obtain precipitation echoes;
step 35: and for the precipitation echo, using the equivalent reflectivity of the 0-y layer as the processed radar echo data.
In the step 3, the radar echo extrapolation model adopts a network structure of ST-LSTM-RNN.
The ST-LSTM-RNN network structure comprises an ST-LSTM unit, a time memory module and a space memory module;
inputting the preprocessed radar echo data into a plurality of ST-LSTM units which are connected in sequence, and outputting predicted radar echo data;
the time memory module and the space memory module are added into each ST-LSTM unit, the time memory module is used for memorizing and accumulating the first m moments of the neurons in the current layer, and the space memory module is used for memorizing and superposing the first m moments of the neurons in different layers.
The time memory module and the space memory module respectively comprise an input gate, an output gate and a forgetting gate;
an input gate for deciding what the input at time t needs to be memorized;
the output gate is used for determining the content of the hidden layer output of the ST-LSTM unit;
the forgetting door is used for controlling the memory content to be forgotten;
and fusing the contents of the time memory module and the space memory module to obtain the final hidden layer output content of the ST-LSTM unit.
Example 1
This embodiment is used in step 2, and the solution of the Z-R relationship is specifically described (taking the area in gansu as an example).
The method mainly comprises the following steps:
step 21: processing the radar echo data, including analysis, quality control and effective elevation selection;
step 22: analyzing an equivalent reflectivity factor from the processed radar echo;
step 23: time weighted averaging is performed on the equivalent reflectance factor per x minutes using the formula:
wherein N represents the number of observation data of the radar within 1 hour, N represents the total number of observation data of the radar within 1 hour, and t1,t2,...,tNRespectively representing the time at which the radar measures the point, Z1,Z2,...,ZNRespectively measuring the 1h average equivalent reflectivity factor value of the point by radar;
step 24: converting the equivalent reflectivity factor after the time weighted average into a reflectivity factor, wherein the adopted formula is as follows:
Ze＝10*lgZ (8)，
where Z represents the reflectivity factor and Ze represents the equivalent reflectivity factor after time-weighted averaging.
In the step 2, the specific steps for obtaining the Z-R relationship are as follows:
step 25: obtaining a Z-R relation by utilizing the reflectivity factor and the precipitation, wherein the Z-R relation is specifically as follows:
Z＝ARb(9)，
wherein Z represents a reflectivity factor, A and b are both parameters, and R represents precipitation;
step 26: and fitting the Z-R relation by using an SPSS tool to obtain the final Z-R relation.
The radar data come from radar data in a user terminal system, and the basic reflectivity (R) is obtained through collection, analysis, quality control, effective elevation selection and the like. Radar data of weather processes with obvious short-time strong precipitation in 2018, 7 and 18 days, 19 days and 20 days of Gansu province and actual 1-hour precipitation observed by corresponding ground stations are used as basic data of Z-R relation research. A total of 2 thousand pieces of data were fitted to the Z-R relationship.
Processing the data includes:
① radar data quality control
And eliminating automatic stations with station numbers not in accordance with longitude and latitude information and rainfall always being zero in the radar coverage range.
The radar data quality control mainly comprises the steps of ground object echo and noise elimination, the identification of bright bands of a brightness layer and the selection of effective elevation angles of the radar. And the mixed scanning reflectivity is calculated by using a climate mixed scanning method, so that the influence of terrain shielding on precipitation estimation is reduced. The time-by-time average reflectivity is obtained by using a time weight averaging method, the randomness of the radar initial value field is overcome, and the method has better representativeness and is closer to the actual distribution of precipitation.
② Radar initial field formation
Since the time of each radar volume sweep is between 5-6min, in order to obtain the average reflectivity of 1h, the embodiment adopts a time-weighted average method to derive an average reflectivity calculation formula per hour.
When there is one observation within 1h, the average reflectivity per hour is calculated as follows:
Z＝Z1(10)，
when there are two observations within 1h, the average reflectivity per hour is calculated as follows:
when three observations are made within 1h, the average reflectance per hour is calculated as follows:
when there are n observations within 1h, the average reflectivity per hour is calculated as follows:
the 1h average reflectivity is calculated by a time weight averaging method, and is synthesized by the time weights of the reflectivities observed by all the radars in the hour.
Analysis of results
① correlation analysis
The SPSS tool is adopted to fit the Z-R relation, the reflectivity factor Z and the rainfall intensity R have strong correlation in a large amount of experience, and the correlation analysis in the SPSS is used for making the correlation between the reflectivity factor Z and the rainfall intensity R in the Gansu area. The results obtained are shown in Table 1:
table 1: correlation test
From the table above, it can be seen that the radar reflectivity factor Z and the rainfall intensity R in the gansu area have a high correlation, and the correlation is 92.1%.
② fitting Z-R relationships
Fitting was performed using the nonlinear fitting method in SPSS, and the fitting results are shown in tables 2-4 and fig. 2:
table 2: description of fit
It can be seen from table 1 that the effect of using a non-linear fit is very good, 92.8% of the data in the training set can be interpreted by the fit function, and the standard error of the estimate is 0.352.
Table 3: analysis of variance (ANOVA)
The significance value sig. corresponding to the F value in table 2 is less than 0.05, which indicates that the overall test effect of the fitting equation is significant, and the fitting equation is useful for practical business problems.
Table 4: coefficient of fit
The parameter a of the fit function obtained by the fitting was 204 and b was 1.61. The errors of parameters a and b are 9.903 and 0.025, respectively, and the effect of coefficient fitting is significant.
As shown in the figure, the fitting effect is significant as shown in the fitted Z-R effect graph, and the final fitting function has parameters of a being 204, b being 1.61, and goodness of fit being 92.8%.
③ Z-R relationship verification
Selecting a Longnan Kangnan forest station and a Linxia Baichuan village station, and respectively performing reverse calculation precipitation comparison of actual precipitation and Z-R relation of the two stations, wherein the comparison results are shown in fig. 3 and 4:
the comparison shows that the variation trend of the back calculation precipitation is consistent with that of the actual precipitation, the variation is larger when sudden strong precipitation occurs, and the back calculation effect of the small precipitation is obvious.
The short-term strong precipitation reflectivity threshold obtained by substituting the level boundary value of the short-term strong precipitation in the Gansu area into the Z-R relational expression by using the Z-R relational expression of the Gansu area obtained above is shown in Table 5:
table 5: short intensity level threshold
The Z-R relation of Gansu integer is Z204R1.61And dividing according to the local short-time strong precipitation standard of Gansu province to obtain the threshold value of the short-time strong precipitation:
table 6: threshold of short-term heavy precipitation
Example 2
This embodiment is used to illustrate step 3,
in step 3, the preprocessing of the radar echo data includes the following steps:
step 31: inputting the radar echo data into a singular point filter and a bilateral filter:
step 32: inputting the radar echo data obtained in the step 31 into an equivalent reflectivity vertical decreasing rate filter, and filtering ground object echoes and partial super-refraction echoes;
step 33: calculating annual average equivalent reflectivity factor distribution by using the radar echo data obtained in the step 32, determining the position shielded by the radar azimuth angle, and filling the related azimuth angle by linear interpolation;
step 34: constructing a high-pass filter to filter the radar echo data obtained in the step 33 to obtain precipitation echoes;
step 35: and for the precipitation echo, using the equivalent reflectivity of the 0-y layer as the processed radar echo data.
In the step 3, the radar echo extrapolation model adopts a network structure of ST-LSTM-RNN.
The ST-LSTM-RNN network structure comprises an ST-LSTM unit, a time memory module and a space memory module;
inputting the preprocessed radar echo data into a plurality of ST-LSTM units which are connected in sequence, and outputting predicted radar echo data;
the time memory module and the space memory module are added into each ST-LSTM unit, the time memory module is used for memorizing and accumulating the first m moments of the neurons in the current layer, and the space memory module is used for memorizing and superposing the first m moments of the neurons in different layers.
The time memory module and the space memory module respectively comprise an input gate, an output gate and a forgetting gate;
an input gate for deciding what the input at time t needs to be memorized;
the output gate is used for determining the content of the hidden layer output of the ST-LSTM unit;
the forgetting door is used for controlling the memory content to be forgotten;
and fusing the contents of the time memory module and the space memory module to obtain the final hidden layer output content of the ST-LSTM unit.
PredRNN networks
Radar nowcasting is essentially a prediction problem of time sequence data, and a neural network needs to predict the position of a radar echo in a future period of time according to the distribution rule of the radar echoes at different times. The key point is that the neural network at different times can obtain the memory of part of the preceding neural network, which is a problem solved by a typical Recurrent Neural Network (RNN).
And (3) carrying out full-field radar echo approach prediction by taking grid points as the minimum prediction unit, and predicting grid points by grid points. In the actual weather process, the radar echo is driven by a weather system to generate development change, so that the echo time change rule of the lattice point and the weather situation around the lattice point need to be considered when the radar echo is predicted, and the radar echo is expressed as the echo characteristic around the lattice point in radar observation. This makes it necessary to take into account not only the storage of temporal information but also the storage of spatial information when constructing the RNN.
In the present embodiment, a network structure of ST-LSTM-RNN is adopted, as shown in fig. 5, compared with a conventional multi-layer RNN architecture, the ST-LSTM-RNN is added to a loop circuit (shown by a dotted line in fig. 5) of the spatial memory module (M) on the basis of a loop of the temporal memory module (C) (shown by a dotted line in fig. 5), so as to enhance propagation of spatial information in neurons at different levels and at different times. In radar nowcasting, the structure is more beneficial to enabling the model to learn radar echo characteristics of different scales and development evolution rules of the radar echo characteristics on a time line. In FIG. 5, W represents an ST-LSTM cell.
FIG. 6 shows the internal structure of the ST-LSTM cell. Ct represents the time memory module in LSTM, which is the memory accumulation at the previous n-th time of the current layer neuron, as shown by the dashed-dotted arrow in FIG. 6. Mt denotes the spatial memory module, which is the memory accumulation of the first n moments of the different levels of neurons, as indicated by the dashed arrows in fig. 6. Ht represents the hidden layer output of the ST-LSTM unit. Similar to the LSTM, the two memory modules each have three respective control gates: the memory content needing to be forgotten in the control module of the forgetting gate (fct, fmt); the input gate (ict, imt) determines what the input at time t needs to be remembered by the module; the output gates (oct, omt) determine the hidden layer output content. And finally, fusing the contents of the two memory modules into a hidden layer of the unit to output Ht.
In radar echo nowcasting, the advantages of the ST-LSTM unit compared with the conventional LSTM unit are mainly embodied in two points:
(1) and the operation process of the state accumulation (ct- > ct +1) and the hidden layer output (ht- > ht +1) is replaced by a convolution form through feedforward calculation. The core is essentially the same as LSTM, with the output of the previous layer being the input to the next layer. The difference is that after the convolution operation is added, the neuron not only can obtain the time sequence relation, but also can extract the spatial feature like the convolution layer.
(2) And a space memory module is added, the data flow of the space memory module is shown by a dotted line in FIG. 6, and the space characteristic information of different dimensions is facilitated to be propagated in the RNN.
And modeling the Guangzhou and Beijing Daxing radars respectively by taking ST-LSTM-RNN as an architecture. Where a set of test subjects included 10 consecutive epochs of radar observations, the model gave 10 consecutive epochs of echo predictions in the future. The model is carried out in a supervised learning mode, and the predicted true value is the radar observation of the next 10 times. The cost function for model convergence is the sum of the root mean square errors of the whole field grid points.
2. Radar data preprocessing
The project uses radar combination reflectivity products to carry out the approach prediction test. The radar observation is influenced by the atmospheric environment, the hardware performance and the like, and the observation result may contain noise generated by a non-meteorological target object, echo pulsation caused by atmospheric turbulence, interference echo and the like, and has a large influence on the forecast result. Therefore, radar data needs to be preprocessed to reduce the influence of clutter. The processing procedure (as shown in fig. 7) is specifically as follows:
(1) the singular point filter and the bilateral filter are constructed to carry out filtering in a value domain and a space domain, and pulsation and clutter can be effectively eliminated on the premise of keeping echo characteristics.
(2) And constructing a reflectivity vertical decreasing rate filter, and filtering the ground object echo and the partial super-refraction echo.
(3) And calculating the annual average reflectivity distribution, determining the position shielded by the azimuth angle, and filling the linear interpolation of the related azimuth angle.
(4) A high pass filter is constructed to remove echoes below 15dBz, leaving only precipitation echoes.
(5) To avoid interference of the bright band of the zero degree layer, only the 0 th-5 th layer reflectivity factor is used in synthesizing the combined reflectivity product.
Fig. 8 shows the result of the preprocessing, fig. 8a shows the combined reflectivity product before preprocessing, fig. 8b shows the combined reflectivity product after preprocessing, in which the clutter at the red circle is effectively suppressed and the radial attenuation in the southwest direction of the radar station is effectively filled. Meanwhile, the form of the echo is basically completely preserved.
3. predRNN model test-long term test
This example compares the PredRNN method (hereinafter referred to as intelligent extrapolation) with the cross-correlation method, which is an algorithm integrated in SWAN2.0 for promotion. Selecting a hit rate (POD), a False Alarm Rate (FAR) and a Critical Success Index (CSI) to carry out quantitative evaluation on a forecast result, and giving a skill score (E) of an intelligent extrapolation method by taking a cross-correlation method as a reference, wherein the calculation method comprises the following steps:
E＝CSIPredRNN-CSICOTREC(14)。
when the number of hits, the number of empty reports and the number of missed reports are calculated, a calculation mode of grid points by grid points is adopted, namely a forecast value and an observed value of the same grid point are selected for comparison. When the observation data as the true value is processed, the same data preprocessing method as the prediction test is used.
The resolution of the forecast product is 0.01 degrees multiplied by 0.01 degrees, the forecast time step is 6min, the forecast aging is 60min at the longest, namely the strength and the position of the combined reflectivity factor after 6min, 12min, 18min, 24min, 30min, 36min, 42min, 48min, 54min and 60min are forecast. By adopting the sub-prediction aging and sub-threshold detection methods, the number of prediction aging is 10, the number of thresholds is 20dBZ, 30dBZ and 50dBZ, and 30 groups of detection results are obtained by detecting grid points with combined reflectivity factors not less than the threshold.
Tables 7-9 show that ① intelligent extrapolation method in the test of two radars has 3 detection thresholds and CSI score higher than that of cross correlation method, ② intelligent extrapolation method in the test of two radars has 3 detection thresholds and POD score higher than that of cross correlation method, FAR lower than that of cross correlation method, ③ intelligent extrapolation method and cross correlation method have prediction capabilities decreasing with the increase of prediction time, specifically, CSI and POD decrease with the increase of time, FAR increases with the increase of time, ④ intelligent extrapolation method and cross correlation method have prediction capabilities decreasing with the increase of combined reflection factor strength, and prediction capabilities are insufficient for the area with strength exceeding 50 dBZ.
Table 7: comparison test result of Beijing Daxing radar test set
Table 8: comparison test result of Beijing Daxing radar test set
Table 9: contrast test result of Guangzhou radar test set
FIG. 9 shows the variation of skill scores of three stations with forecast age, respectively, where the skill score is greater than 0, indicating that the forecast capacity of the intelligent extrapolation exceeds the cross-correlation method, from tables 7-9, it can be seen that ① the intelligent extrapolation method has a CSI score exceeding the cross-correlation method in the two radar trials in all the test items, ② has a skill score increasing with the rise of the forecast age in the 20dBZ and 30dBZ test items, indicating that the forecast capacity of the intelligent extrapolation method decreases more slowly with the rise of the forecast age in the two reflectivity intervals, ③ has a relatively lowest skill score in the 50dBZ test item, but considering that the CSI scores of the two methods in this interval are both low, a skill score of 0.05 also brings about a significant improvement in the skill score
Example 3
This embodiment will explain the present invention by taking a langzhou radar as an example.
From 08 days 02 to 08 days 03, there are small rains in most parts of Gansu province or 5 heavy rainstorms in summer and Lanzhou city in the weather of showering.
Fig. 10 shows a comparison graph of the situation of the intelligent extrapolation prediction, and fig. 10 is a dynamic change graph of the situation of the intelligent extrapolation result in 1 hour in the forecast of the situation of 00:54 in 8 months and 3 days in 2018, 8 months and 3 days in 01:54 in 2018, 8 months and 3 days in 2018, and 1 hour in 8 months and 3 days in 2018, respectively. And the comparison shows that the moving direction speed of the whole echo zone predicted by extrapolation is basically consistent with the live condition, and the intensity and the range are also equivalent to the live condition. Wherein, the echo of more than 30dBZ in the northwest direction of the Lanzhou radar station is strengthened, and the range is enlarged. The 30dBZ in the due north direction reduces the echo range, and the echo has a dissipation trend. The forecast is basically accurate.
Performing back-stepping on echo data of Qingyang radar of 30/8 in 2018 by using an intelligent extrapolation algorithm, analyzing a result (geometric correction, radiation correction and the like), performing longitude and latitude correspondence with a meteorological site covered by the Qingyang radar (the distance deviation between the longitude and latitude is less than 0.5KM), performing time-weighted cumulative averaging within 1h on echo reflectivity values obtained by the back-stepping of the radar echo in every 6 minutes after the correspondence is performed, converting the weighted cumulative average of the echo reflectivity into a reflectivity factor value Z by using a formula, and combining a mixed cloud Z-R relational expression Z-212R in Gansu1.35And (3) calculating the hourly precipitation amount R of the corresponding site, and finally comparing and analyzing the hourly precipitation amount R with the live precipitation amount of the site, wherein the result is obtained by converting a back-thrust result 1h before 18 o' clock in 8 and 18 th of 2018 of Qingyang radar into precipitation amount and comparing the precipitation amount with the live precipitation amount of the site as shown in fig. 11.
Example 4
This example is used to illustrate the selection of the effective elevation angle of the radar in example 1.
In the aspect of researching the use of a Doppler radar for forecasting precipitation, the selection of the radar elevation angle is a ubiquitous and important factor. The method and the device find the relation between elevation of grid points in Gansu province and radar stations to solve the problem of selecting elevation angles of the radar in the forecast precipitation of different stations.
According to the longitude and latitude and the altitude of a central point (more than 6000 and ten thousand total grid point data) divided by grid points with 30 × 30m spatial resolution in the whole world of Gansu province, the selection of the forecast radar elevation of each elevation grid point in each radar range when rainfall is forecasted is calculated based on the longitude and latitude, the altitude and the scanning range of 7 radars (Jiayuguan, Zhangye, Lanzhou, Gannan, Longnan, Tianshui and Xifeng) in Gansu province.
The selection method of the elevation angle comprises the following steps:
(1) calculating the earth surface distance between each elevation lattice point and the radar, wherein the lattice points are square, so that the center points of the lattice points are used as the basis for calculation;
(2) acquiring a grid point range scanned by the radar by using the scanning range of the radar;
(3) calculating a ground surface included angle formed by each grid point and a radar in the grid point range by using the ground surface distance; first, calculating cos value of the earth surface included angle, thereby obtaining the final earth surface included angle.
(4) And obtaining the effective elevation angle of the radar by utilizing the ground surface included angle.
(41) Screening all grid points within a range of +/-0.5 degrees of the earth surface included angle corresponding to any grid point a in the grid points as a data set A;
(42) further screening the grid points in the data set A to obtain grid points with the earth surface distance not greater than the earth surface distance between the grid points a and the radar, and using the grid points as a data set B;
(43) for the earth surface included angle corresponding to the grid point in the data set B, taking the maximum value of the earth surface included angle as the elevation angle of the radar scanning grid point a;
(44) and obtaining the effective elevation angle of the radar scanning the lattice point a according to the elevation angle and the actual elevation angle of the radar, wherein the actual elevation angle of the radar is as follows: 0.5, 1.5, 2.4, 3.4, 4.3, 6.0, so the closest elevation is taken as the actual effective elevation for that grid point.
The method comprises the steps of selecting a 100KM area covered by a Longnan radar station (Z9939) to perform radar elevation angle selection calculation based on DEM elevation to obtain the following graph (an actual elevation map of the Longnan radar coverage area is shown in figure 12, and a radar elevation angle selection map of the Longnan radar coverage area is shown in figure 13).
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (7)
1. A radar echo extrapolation short-term prediction method based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
step 1: acquiring actually measured radar echo data and rainfall;
step 2: obtaining an equivalent reflectivity factor by using the radar echo data, and obtaining a Z-R relation by using the reflectivity factor and rainfall;
and step 3: constructing a radar echo extrapolation model, preprocessing the radar echo data, training the radar echo extrapolation model by using the processed data, and predicting the radar echo data in a future period of time;
and 4, step 4: and inputting the predicted radar echo data into a Z-R relation to obtain the predicted rainfall.
2. The method of claim 1, wherein the method comprises: in the step 2, the step of obtaining the reflectivity factor specifically includes:
step 21: processing the radar echo data, including analysis, quality control and effective elevation selection;
step 22: analyzing an equivalent reflectivity factor from the processed radar echo;
step 23: time weighted averaging is performed on the equivalent reflectance factor per x minutes using the formula:
wherein N represents the number of observation data of the radar within 1 hour, N represents the total number of observation data of the radar within 1 hour, and t1,t2,...,tNRespectively representing the time at which the radar measures the point, Z1,Z2,...,ZNRespectively measuring the 1h average equivalent reflectivity factor value of the point by radar;
step 24: converting the equivalent reflectivity factor after the time weighted average into a reflectivity factor, wherein the adopted formula is as follows:
Ze＝10*lgZ(2)，
where Z represents the reflectivity factor and Ze represents the equivalent reflectivity factor after time-weighted averaging.
3. The method of claim 2, wherein the method comprises: in the step 2, the specific steps for obtaining the Z-R relationship are as follows:
step 25: obtaining a Z-R relation by utilizing the reflectivity factor and the precipitation, wherein the Z-R relation is specifically as follows:
Z＝ARb(3)，
wherein Z represents a reflectivity factor, A and b are both parameters, and R represents precipitation;
step 26: and fitting the Z-R relation by using an SPSS tool to obtain the final Z-R relation.
4. The method of claim 1, wherein the method comprises: in the step 3, the preprocessing of the radar echo data comprises the following steps
Step 31: inputting the radar echo data into a singular point filter and a bilateral filter:
step 32: inputting the radar echo data obtained in the step 31 into an equivalent reflectivity vertical decreasing rate filter, and filtering ground object echoes and partial super-refraction echoes;
step 33: calculating annual average equivalent reflectivity factor distribution by using the radar echo data obtained in the step 32, determining the position shielded by the radar azimuth angle, and filling the related azimuth angle by linear interpolation;
step 34: constructing a high-pass filter to filter the radar echo data obtained in the step 33 to obtain precipitation echoes;
step 35: and for the precipitation echo, using the equivalent reflectivity of the 0-y layer as the processed radar echo data.
5. The method of claim 1, wherein the method comprises: in the step 3, the radar echo extrapolation model adopts a network structure of ST-LSTM-RNN.
6. The method of claim 5, wherein the method comprises: the ST-LSTM-RNN network structure comprises an ST-LSTM unit, a time memory module and a space memory module;
inputting the preprocessed radar echo data into a plurality of ST-LSTM units which are connected in sequence, and outputting predicted radar echo data;
the time memory module and the space memory module are added into each ST-LSTM unit, the time memory module is used for memorizing and accumulating the first m moments of the neurons in the current layer, and the space memory module is used for memorizing and superposing the first m moments of the neurons in different layers.
7. The method of claim 6, wherein the method comprises: the time memory module and the space memory module respectively comprise an input gate, an output gate and a forgetting gate;
an input gate for deciding what the input at time t needs to be memorized;
the output gate is used for determining the content of the hidden layer output of the ST-LSTM unit;
the forgetting door is used for controlling the memory content to be forgotten;
and fusing the contents of the time memory module and the space memory module to obtain the final hidden layer output content of the ST-LSTM unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911029867.9A CN110967695A (en) | 2019-10-28 | 2019-10-28 | Radar echo extrapolation short-term prediction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911029867.9A CN110967695A (en) | 2019-10-28 | 2019-10-28 | Radar echo extrapolation short-term prediction method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110967695A true CN110967695A (en) | 2020-04-07 |
Family
ID=70029894
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911029867.9A Pending CN110967695A (en) | 2019-10-28 | 2019-10-28 | Radar echo extrapolation short-term prediction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110967695A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200349A (en) * | 2020-09-16 | 2021-01-08 | 平衡机器科技(深圳)有限公司 | Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN |
| CN112232674A (en) * | 2020-10-16 | 2021-01-15 | 中国气象局气象探测中心 | Meteorological disaster assessment method, device and system |
| CN112698427A (en) * | 2020-12-09 | 2021-04-23 | 最美天气(上海)科技有限公司 | Short-term forecasting method and system based on space-time forecasting model |
| CN112946652A (en) * | 2021-01-21 | 2021-06-11 | 中国气象科学研究院 | Method and device for filling radar beam occlusion area |
| CN113253275A (en) * | 2021-04-22 | 2021-08-13 | 南京航空航天大学 | Rainfall estimation method based on improved RBF neural network |
| CN113267834A (en) * | 2020-11-30 | 2021-08-17 | 武汉超碟科技有限公司 | Fusion rainfall forecasting method based on multi-model integration |
| CN113640803A (en) * | 2021-09-01 | 2021-11-12 | 江西师范大学 | Short-time quantitative rainfall forecasting method based on echo intensity and echo top height extrapolation |
| CN115453541A (en) * | 2022-11-14 | 2022-12-09 | 中科星图维天信(北京)科技有限公司 | Precipitation amount prediction method and device, electronic device and storage medium |
| CN115792847A (en) * | 2022-11-08 | 2023-03-14 | 江西师范大学 | Quantitative precipitation estimation method based on neural network and echo vertical information |
| CN117452369A (en) * | 2023-12-25 | 2024-01-26 | 江西师范大学 | Echo jacking calculation optimization method for short-time disastrous weather monitoring |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106526708A (en) * | 2016-09-21 | 2017-03-22 | 广东奥博信息产业有限公司 | Intelligent early-warning analysis method for meteorological severe convection weather based on machine learning |
| CN106872981A (en) * | 2017-02-17 | 2017-06-20 | 水利部南京水利水文自动化研究所 | The precipitation strong center tracking of rainfall radar and forecasting procedure |
| CN108508505A (en) * | 2018-02-05 | 2018-09-07 | 南京云思创智信息科技有限公司 | Heavy showers and thunderstorm forecasting procedure based on multiple dimensioned convolutional neural networks and system |
| CN108983323A (en) * | 2018-08-08 | 2018-12-11 | 湖北河海科技发展有限公司 | Precipitation forecast method and early warning platform based on optical flow method |
| CN109598052A (en) * | 2018-11-29 | 2019-04-09 | 武汉大学 | Intelligent electric meter life cycle prediction technique and device based on correlation analysis |
| CN109814175A (en) * | 2019-02-14 | 2019-05-28 | 浙江省气象台 | A kind of satellite-based strong convection monitoring method and its application |
-
2019
- 2019-10-28 CN CN201911029867.9A patent/CN110967695A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106526708A (en) * | 2016-09-21 | 2017-03-22 | 广东奥博信息产业有限公司 | Intelligent early-warning analysis method for meteorological severe convection weather based on machine learning |
| CN106872981A (en) * | 2017-02-17 | 2017-06-20 | 水利部南京水利水文自动化研究所 | The precipitation strong center tracking of rainfall radar and forecasting procedure |
| CN108508505A (en) * | 2018-02-05 | 2018-09-07 | 南京云思创智信息科技有限公司 | Heavy showers and thunderstorm forecasting procedure based on multiple dimensioned convolutional neural networks and system |
| CN108983323A (en) * | 2018-08-08 | 2018-12-11 | 湖北河海科技发展有限公司 | Precipitation forecast method and early warning platform based on optical flow method |
| CN109598052A (en) * | 2018-11-29 | 2019-04-09 | 武汉大学 | Intelligent electric meter life cycle prediction technique and device based on correlation analysis |
| CN109814175A (en) * | 2019-02-14 | 2019-05-28 | 浙江省气象台 | A kind of satellite-based strong convection monitoring method and its application |
Non-Patent Citations (2)
| Title |
|---|
| 韩丰: "循环神经网络在雷达临近预报中的应用", 《应用气象学报》 * |
| 魏明英: "基于多普勒雷达资料的降水定量估测及其应用研究", 《中国优秀硕士学位论文全文数据库基础科学辑》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200349A (en) * | 2020-09-16 | 2021-01-08 | 平衡机器科技(深圳)有限公司 | Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN |
| CN112200349B (en) * | 2020-09-16 | 2022-06-03 | 平衡机器科技(深圳)有限公司 | Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN |
| CN112232674A (en) * | 2020-10-16 | 2021-01-15 | 中国气象局气象探测中心 | Meteorological disaster assessment method, device and system |
| CN112232674B (en) * | 2020-10-16 | 2021-12-07 | 中国气象局气象探测中心 | Meteorological disaster assessment method, device and system |
| CN113267834A (en) * | 2020-11-30 | 2021-08-17 | 武汉超碟科技有限公司 | Fusion rainfall forecasting method based on multi-model integration |
| CN112698427A (en) * | 2020-12-09 | 2021-04-23 | 最美天气(上海)科技有限公司 | Short-term forecasting method and system based on space-time forecasting model |
| CN112946652B (en) * | 2021-01-21 | 2023-03-28 | 中国气象科学研究院 | Method and device for filling radar beam occlusion area |
| CN112946652A (en) * | 2021-01-21 | 2021-06-11 | 中国气象科学研究院 | Method and device for filling radar beam occlusion area |
| CN113253275A (en) * | 2021-04-22 | 2021-08-13 | 南京航空航天大学 | Rainfall estimation method based on improved RBF neural network |
| CN113640803A (en) * | 2021-09-01 | 2021-11-12 | 江西师范大学 | Short-time quantitative rainfall forecasting method based on echo intensity and echo top height extrapolation |
| CN115792847A (en) * | 2022-11-08 | 2023-03-14 | 江西师范大学 | Quantitative precipitation estimation method based on neural network and echo vertical information |
| CN115453541A (en) * | 2022-11-14 | 2022-12-09 | 中科星图维天信(北京)科技有限公司 | Precipitation amount prediction method and device, electronic device and storage medium |
| CN117452369A (en) * | 2023-12-25 | 2024-01-26 | 江西师范大学 | Echo jacking calculation optimization method for short-time disastrous weather monitoring |
| CN117452369B (en) * | 2023-12-25 | 2024-04-05 | 江西师范大学 | Echo jacking calculation optimization method for short-time disastrous weather monitoring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110967695A (en) | Radar echo extrapolation short-term prediction method based on deep learning | |
| Bianco et al. | Convective boundary layer depth: Improved measurement by Doppler radar wind profiler using fuzzy logic methods | |
| Anagnostou | A convective/stratiform precipitation classification algorithm for volume scanning weather radar observations | |
| Goudenhoofdt et al. | Generation and verification of rainfall estimates from 10-yr volumetric weather radar measurements | |
| CN110346844A (en) | Quantitative Precipitation estimating and measuring method of the NRIET based on cloud classification and machine learning | |
| CN100510774C (en) | Synthetic aperture radar image self-adaptive spot noise suppressing method | |
| Liberman et al. | New algorithm for integration between wireless microwave sensor network and radar for improved rainfall measurement and mapping | |
| CN111289983B (en) | Inversion method for vertically accumulated liquid water content of radar | |
| CN104569981A (en) | Synergy self-adaption observing method | |
| CN110261857B (en) | Spatial interpolation method for weather radar | |
| Nai et al. | On the mitigation of wind turbine clutter for weather radars using range‐Doppler spectral processing | |
| CN113311416A (en) | Mountain region small watershed radar quantitative precipitation estimation technology | |
| CN113933845A (en) | Ground hail reduction identification and early warning method based on dual-linear polarization radar | |
| CN115166750A (en) | Quantitative precipitation estimation method based on dual-polarization Doppler radar data | |
| CN113238230B (en) | Strong wind early warning method for power grid production caused by strong convection in summer | |
| Lee et al. | Mixing depth estimation from operational JMA and KMA wind-profiler data and its preliminary applications: Examples from four selected sites | |
| Williams et al. | Cluster analysis techniques to separate air motion and hydrometeors in vertical incident profiler observations | |
| CN113219463A (en) | Radar ground object echo identification method and system for power system | |
| Hirano et al. | Method of VIL calculation for X-band polarimetric radar and potential of VIL for nowcasting of localized severe rainfall-Case study of the Zoshigaya downpour, 5 August 2008 | |
| Kato et al. | Very short time range forecasting using CReSS-3DVAR for a meso-γ-scale, localized, extremely heavy rainfall event: Comparison with an extrapolation-based nowcast | |
| Einfalt et al. | Quality control of precipitation data | |
| CN114935759A (en) | Wave field missing value filling method and system based on high-frequency ground wave radar observation | |
| Cui et al. | Association of lightning occurrence with precipitation cloud column structure at a fixed position | |
| Ackermann et al. | Radar and environment-based hail damage estimates using machine learning | |
| Pérez Hortal et al. | Using a continental-scale data quality monitoring framework to evaluate a new nonweather filter for radar observations |
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: 20200407 |
|
| RJ01 | Rejection of invention patent application after publication |