A method for calculating radar vertical mapping error in urban environment

Abstract

The application provides a radar vertical mapping error calculation method in an urban environment, and relates to the technical field of meteorological analysis. The method comprises the following steps: based on a radar reflectivity dataset of a target city, a corresponding precipitation inversion result of each height layer is calculated through a pre-constructed layered radar precipitation inversion model. The input of the layered radar precipitation inversion model is radar reflectivity data of the corresponding height layer, and the output is the precipitation inversion result. Based on the driving factor dataset of each height layer in the target city, the height error characteristics of each height layer are calculated through a pre-constructed height error characteristic mapping function. The driving factor dataset of the target city, the precipitation inversion result and the height error characteristics of each height layer are input into a pre-trained surface precipitation prediction model for prediction processing to obtain the final surface precipitation inversion result of the target city. The application aims to improve the accuracy of surface precipitation inversion.

Description

Technical Field

[0001] This application relates to the field of meteorological analysis technology, specifically to a method for calculating radar vertical mapping error in an urban environment. Background Technology

[0002] Against the backdrop of global climate change and rapid urbanization, the frequency and intensity of extreme precipitation events have increased significantly, exacerbating the risk of urban rainstorm disasters. In this context, weather radar, with its wide-area and continuous observation capabilities, is widely used for quantitative precipitation estimation, serving as a crucial foundation for urban flood control and drainage, disaster early warning, and refined risk assessment.

[0003] Existing radar precipitation retrieval and vertical mapping methods mostly employ empirical extrapolation or simple correction to map upper-air radar observations to surface precipitation. However, in practical applications, current schemes often focus on the correspondence between single-height-level retrieval results and surface precipitation, leading to insufficient vertical mapping accuracy in complex urban environments. Furthermore, radar vertical mapping errors in urban environments are influenced by various meteorological conditions and underlying surface factors, such as wind field, humidity structure, air pressure changes, and vegetation distribution. Current schemes fail to effectively incorporate multi-source meteorological and underlying surface information, resulting in insufficient explanatory power for the causes of errors. Summary of the Invention

[0004] In view of this, this application provides a method for calculating radar vertical mapping error in an urban environment. It aims to solve or partially solve the problems existing in the background technology.

[0005] The first aspect of this application provides a method for calculating radar vertical mapping error in an urban environment, the method comprising: Based on the radar reflectivity dataset of the target city, the precipitation inversion results corresponding to each height layer are calculated through pre-constructed layered radar precipitation inversion models. The input of the layered radar precipitation inversion model is the radar reflectivity data of the corresponding height layer, and the output is the precipitation inversion result. Based on the driving factor dataset of the i-th altitude layer in the target city, the altitude error features of the i-th altitude layer are calculated through a pre-constructed altitude error feature mapping function. The target city's driving factor dataset, precipitation inversion results for each altitude layer, and altitude error features are input into a pre-trained surface precipitation prediction model for prediction processing to obtain the final surface precipitation inversion results for the target city. The driving factors include atmospheric dynamic driving factors, atmospheric thermal driving factors, and underlying surface driving factors. The height error feature of the i-th height layer is an N-dimensional array corresponding to the i-th height layer. The j-th element of the N-dimensional array is the vertical mapping error vector between the i-th and j-th height layers. The vertical mapping error vector includes a correlation index, an amplitude deviation index, and a random error index between the two corresponding height layers, where i and j range from 1 to N, and N is the total number of height layers. The correlation index is an indicator of the consistency of the precipitation inversion results of the two corresponding height layers in terms of spatial distribution structure. The amplitude deviation index is an indicator of the systematic difference in amplitude of the precipitation inversion results of the two corresponding height layers. The random error index is a random error index of the precipitation inversion results of the two corresponding height layers at the local scale.

[0006] The radar vertical mapping error calculation method provided in this application for an urban environment has the following advantages: The radar vertical mapping error calculation method provided in this application for urban environments achieves effective integration of the advantageous information from radar reflectivity data at different altitudes for surface precipitation retrieval under various meteorological conditions by nonlinearly and collaboratively fusing the retrieval results of surface precipitation from radar reflectivity data at different altitudes. Simultaneously, the method introduces multi-source meteorological elements and urban underlying surface characteristics, which can influence vertical mapping error, as error-driving factors during the fusion process to correct the vertical mapping error. Compared to previous surface precipitation retrieval results based on radar reflectivity data at a single altitude, this method demonstrates higher stability and accuracy in its retrieval results. Attached Figure Description

[0007] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a flowchart illustrating a radar vertical mapping error calculation method in an urban environment, as shown in one embodiment of this application. Figure 2 This is another flowchart illustrating a radar vertical mapping error calculation method in an urban environment, as shown in one embodiment of this application; Figure 3 This is a comparison diagram illustrating the inversion effect of a radar vertical mapping error calculation method in an urban environment, as shown in one embodiment of this application. Detailed Implementation

[0009] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0010] refer to Figure 1 , Figure 1 This is a flowchart illustrating a radar vertical mapping error calculation method in an urban environment, as shown in one embodiment of this application. Figure 1 As shown, the method includes: Step S01: Based on the radar reflectivity dataset of the target city, the precipitation inversion results corresponding to each height layer are calculated through the pre-constructed layered radar precipitation inversion models. The input of the layered radar precipitation inversion model is the radar reflectivity data of the corresponding height layer, and the output is the precipitation inversion result.

[0011] In this embodiment, the present application pre-constructs multiple layered radar precipitation inversion models. Each layered radar precipitation inversion model corresponds to a height layer. One layered radar precipitation inversion model is used to invert precipitation inversion results based on radar reflectivity data of the corresponding height layer.

[0012] For example, two layered radar precipitation inversion models, 1 and 2, were pre-constructed. Layered radar precipitation inversion model 1, based on radar reflectivity data at a height of 500 meters above the surface for any input location, inverts and outputs the precipitation inversion result for that location. Layered radar precipitation inversion model 2, based on radar reflectivity data at a height of 1000 meters above the surface for any input location, inverts and outputs the precipitation inversion result for that location. It should be noted that the precipitation inversion result output by the layered radar precipitation inversion model corresponding to a specific height is the precipitation inversion result at that location at that height. Ultimately, this application obtains the surface precipitation inversion result for the same location based on precipitation inversion results at different heights.

[0013] First, obtain the radar reflectivity dataset of the target city during the current rainfall event. The radar reflectivity data in this dataset is in the following format: Where h represents the altitude, such as 500 meters or 1000 meters above the Earth's surface, t represents the time, and (x, y) represents the spatial position in the plane. This represents the radar reflectivity at the corresponding altitude layer at the corresponding spatial location at the corresponding time.

[0014] For the acquired radar reflectivity dataset, radar reflectivity data are sequentially obtained from the dataset. Based on the height layer in the currently acquired radar reflectivity data, the radar reflectivity data is input into the layered radar precipitation inversion model corresponding to that height layer. The layered radar precipitation inversion model then calculates and outputs the precipitation inversion result for the spatial location corresponding to the radar reflectivity data at the corresponding time.

[0015] For example, the radar reflectivity data currently acquired is the radar reflectivity at a spatial location 'a' at a height of 500 meters above the Earth's surface at time 't'. This radar reflectivity data is then input into the layered radar precipitation inversion model corresponding to this 500-meter height. The layered radar precipitation inversion model then calculates and outputs the precipitation inversion result for spatial location 'a' at time 't'.

[0016] In this embodiment, radar reflectivity data at multiple different altitude levels exist for the same spatial location at the same time. Therefore, by inputting the radar reflectivity data for each altitude level into the corresponding layered radar precipitation inversion model, multiple precipitation inversion results corresponding to different altitude levels for the same spatial location at the same time can be obtained. For example, if spatial location a has radar reflectivity data at distances of 500 meters, 750 meters, and 1000 meters from the ground surface at time t, inputting these three data points into the corresponding layered radar precipitation inversion model will yield three precipitation inversion results corresponding to different altitude levels for spatial location a at time t.

[0017] Using the same implementation method, multiple precipitation inversion results corresponding to the altitude layer will be obtained for each spatial location in the radar reflectivity dataset through inversion calculation.

[0018] In this application, layered radar precipitation inversion models corresponding to each altitude layer are pre-constructed, and the expressions of the layered radar precipitation inversion models are as follows:

[0019] in, Represents the radar reflectivity data of the spatial position (x, y) in the plane at time t in the h-th spatial layer; α and β represent evolution parameters; The precipitation inversion result is obtained by inverting the radar reflectivity data of the plane spatial location (x,y) at time t in the h-th spatial layer.

[0020] In this embodiment, this application first pre-constructs a corresponding layered radar precipitation inversion model for each altitude layer. That is, there is a corresponding layered radar precipitation inversion model for each altitude layer. The number of altitude layers and the distance between them can be set according to the actual application scenario. For example, three altitude layers can be selected, namely, altitude layers at 10 meters, 20 meters, and 30 meters above the ground surface. In each layered radar precipitation inversion model, the evolution parameters α and β are unknowns. Their values ​​are obtained by linearly fitting the radar reflectivity of the 0th layer (i.e., the radar reflectivity of the lowest layer) with historical precipitation data. The values ​​of these two evolution parameters are used in the layered radar precipitation inversion models of other altitude layers.

[0021] Step S02: Based on the driving factor dataset of the i-th altitude layer in the target city, the altitude error features of the i-th altitude layer are calculated through a pre-constructed altitude error feature mapping function. The driving factor types include atmospheric dynamic driving factors, atmospheric thermal driving factors, and underlying surface driving factors. The atmospheric dynamic driving factors include wind speed factors and wind direction factors. The atmospheric thermal driving factors include inlet pressure factors, air temperature factors, and relative humidity factors. The underlying surface driving factors include high vegetation leaf area index, low vegetation leaf area index, and impermeable surface ratio.

[0022] In this embodiment, the driving factor data of the target city in the current rainfall event is obtained, and the obtained driving factor data also corresponds to the altitude layer. Finally, the driving factor dataset of the target city at each altitude layer in the current rainfall event is obtained. The driving factor dataset includes atmospheric dynamic driving factors, atmospheric thermal driving factors, and underlying surface driving factors.

[0023] For example, if the driving factors include wind speed, temperature, and impervious surface ratio, then the target city will obtain corresponding wind speed data, temperature data, and impervious surface ratio data for each altitude level during the current rainfall event. For instance, the wind speed at an altitude level of 500 meters above the ground is a1, the temperature is b1, and the impervious surface ratio is c1; the wind speed at an altitude level of 1000 meters above the ground is a2, the temperature is b2, and the impervious surface ratio is c2.

[0024] In this embodiment, regarding atmospheric dynamic driving factors, this application selects wind speed and wind direction factors to characterize the advection transport and tilting effects of precipitation particles at different altitudes; regarding atmospheric thermal driving factors, this application selects air pressure, air temperature, and relative humidity factors to reflect the modulation effect of atmospheric stability, melt layer height, and evaporation process on the vertical distribution of radar echoes; regarding underlying surface driving factors, this application selects high vegetation leaf area index, low vegetation leaf area index, and impermeable surface ratio factors to describe the influence of urban underlying surfaces on near-surface thermal structure, turbulence intensity, and convection triggering conditions.

[0025] In this embodiment, a height error feature mapping function is pre-created. This function, based on the input driving factor dataset of the i-th height layer, can calculate and output the height error feature corresponding to the i-th height layer, where i ranges from 1 to N, and N is the total number of height layers. The height error feature corresponding to each height layer is an N-dimensional array, where N represents the total number of height layers. The specific composition of the height error feature is explained below: The height error feature corresponding to the i-th height layer is an N-dimensional array. The j-th element of this N-dimensional array is the vertical mapping error vector between the i-th and j-th height layers. This vertical mapping error vector includes a correlation index, an amplitude deviation index, and a random error index between the i-th and j-th height layers. Here, i and j both range from 1 to N, where N is the total number of height layers. The correlation index between the i-th and j-th height layers is an indicator of the consistency of the spatial distribution structure of the precipitation inversion results between these two height layers. The amplitude deviation index between the i-th and j-th height layers is an indicator of the systematic difference in amplitude between the precipitation inversion results of these two height layers. The random error index between the i-th and j-th height layers is a random error index of the precipitation inversion results at the local scale.

[0026] In this embodiment, it should be understood that the relevant data and inversion results in this application are time-dependent. For example, based on the radar reflectivity dataset at time t, the inversion calculation yields the inversion results at time t corresponding to each height layer. Based on the driving factor dataset of the i-th height layer at time t, the height error characteristics of the i-th height layer at time t are calculated using a pre-constructed height error feature mapping function.

[0027] In this application, a height error feature mapping function is constructed, including: Step S02_01: Construct an initial random forest model. The input of the initial random forest model is the height layer and the corresponding driving factor dataset, and the output is the height error feature corresponding to the height layer.

[0028] In this embodiment, an optional implementation method is to construct a height error feature mapping function using a random forest algorithm. Specifically, this application pre-constructs an initial random forest model, the input of which is a height layer and the corresponding driving factor dataset, and the output is the height error feature corresponding to the height layer.

[0029] Step S02_02: Input the driving factor dataset of the i-th height layer of historical rainfall events into the initial random forest model to obtain the predicted height error features of the i-th height layer.

[0030] In this embodiment, the driving factor datasets for each altitude layer of the target city at time t during historical rainfall events are obtained. The i-th altitude layer and its corresponding driving factor dataset at time t are input into the constructed initial random forest model. Based on the input data, the initial random forest model calculates the predicted altitude error features of the i-th altitude layer at time t. Then, the standard altitude error features of the target city at time t during the historical rainfall events are obtained.

[0031] In this application, based on the radar reflectivity dataset of the historical rainfall events, the standard height error characteristics of each altitude layer are calculated, including: Step S02_02a: Based on the radar reflectivity dataset of the historical rainfall events, the historical precipitation inversion results corresponding to each height layer are calculated using the pre-constructed layered radar precipitation inversion models.

[0032] In this embodiment, the radar reflectivity dataset of the target city at time t during the historical rainfall event is obtained. Using pre-constructed layered radar precipitation inversion models, the historical precipitation inversion results at time t corresponding to each altitude layer are calculated. The calculation method is the same as step S01 described above and will not be repeated here.

[0033] Step S02_02b: Based on the historical precipitation inversion results, calculate the vertical mapping error vector between the i-th height layer and the j-th height layer; The expression for calculating the correlation index in the vertical mapping error vector is as follows: ,in, is the correlation index between the i-th and j-th altitude layers; n is the total number of valid spatial sampling points within the target city area; , These represent the precipitation inversion values ​​at the k-th spatial location for the i-th and j-th altitude layers, respectively. , represent the average precipitation inversion values ​​of the i-th and j-th altitude layers within the target city area, respectively; The expression for calculating the magnitude deviation index in the vertical mapping error vector is as follows: ,in, This is the magnitude deviation index between the i-th and j-th height layers; The expression for calculating the random error index in the vertical mapping error vector is as follows: ,in, This is the random error index between the i-th and j-th altitude layers.

[0034] In this embodiment, based on the historical precipitation inversion results at time t corresponding to each height layer obtained in step S02_02a, the values ​​of each index in the vertical mapping error vector between the i-th height layer and the j-th height layer are calculated using the calculation expressions of the correlation index, the amplitude deviation index, and the random error index. The values ​​of i and j are any positive integers from 1 to N, and N is the total number of height layers.

[0035] Step S02_02c: Based on all the calculated vertical mapping error vectors, generate the standard height error features corresponding to each height layer.

[0036] In this embodiment, after calculating the vertical mapping error vectors of any two height layers at time t through step S02_02b, the standard height error characteristics of each height layer at time t are obtained by grouping all the obtained vertical mapping error vectors.

[0037] In this application, the standard height error features required for model training are calculated using the aforementioned three metrics. In the practical application phase, after the model training is complete and the corresponding height error feature mapping function is obtained, the height error features are calculated using this constructed height error feature mapping function, thereby improving computational efficiency.

[0038] Step S02_03: Calculate the residual between the predicted height error feature and the standard height error feature of the same height layer, and iteratively optimize the initial random forest model with the goal of minimizing the sum of all residuals until the model accuracy reaches a preset threshold, thus obtaining a trained random forest model; wherein, the standard height error feature of each height layer is calculated based on the radar reflectivity dataset of the historical rainfall events.

[0039] In this embodiment, after obtaining the predicted height error features and standard height error features of each height layer in the target city at time t during the historical rainfall event, the residuals between the predicted height error features and standard height error features of the same height layer are calculated. After obtaining the residuals corresponding to each height layer, the initial random forest model is iteratively optimized with the objective of minimizing the sum of all residuals until the model accuracy reaches a preset threshold, or when the sum of all residuals tends to stabilize, thus obtaining a successfully trained random forest model.

[0040] Step S02_04: Determine the height error feature mapping function for the completed training of the random forest model.

[0041] In this embodiment, the trained random forest model is ultimately determined as the height error feature mapping function.

[0042] Step S03: Input the target city's driving factor dataset, precipitation inversion results corresponding to each altitude layer, and altitude error features into the pre-trained surface precipitation prediction model for prediction processing to obtain the final surface precipitation inversion results for the target city.

[0043] In this embodiment, the driving factor datasets for each altitude layer of the target city at time t, the precipitation inversion results for each altitude layer at time t, and the altitude error features for each altitude layer at time t are all embedded into a unified temporal feature space to obtain the corresponding input sequence. The expression of this input sequence is as follows: ,in, This represents the precipitation inversion result at time t for the i-th altitude layer; This represents the height error characteristics of the i-th height layer at time t; This represents the value of the m-th driving factor at time t at the i-th altitude layer. Then, this input sequence is input into a pre-trained, directly applicable surface precipitation prediction model for prediction processing, yielding the final surface precipitation inversion result for the target city at time t. This surface precipitation inversion result is expressed as... , This represents the surface precipitation inversion value at time t for a planar spatial location (x, y).

[0044] The preferred surface precipitation prediction model is an LSTM network architecture. When the preferred surface precipitation prediction model is an LSTM network architecture, an initial LSTM network is constructed to learn the dynamic nonlinear mapping relationship between the radar multi-height precipitation inversion results and the final surface precipitation evolution.

[0045] In an LSTM network, the input sequence Implicit state being progressively passed on to each moment The hidden state at each time step From the implicit state of the previous moment and the input at the current moment Co-updating. A key feature of this LSTM network is its memory units, which enable it to maintain and update historical information, thereby capturing long-term dependencies in time-series data. The LSTM update process is controlled by the following four gating mechanisms: forget gate... It determines the hidden state of the previous moment. How much information needs to be forgotten; input gate It determines the input sequence at the current moment. How much information needs to be stored; candidate memory units It generates candidate memory information that needs to be updated at the current moment; the output gate It determines the memory unit at the current moment. How much information should be output as a hidden state? The specific formula is as follows: Forgotten Gate:

[0046] Input Gate:

[0047] Candidate memory units:

[0048] Memory unit update:

[0049] Output gate:

[0050] Implicit state:

[0051] in, , , , Here is the weight matrix for the corresponding gate. , , , This is the bias term. Using the above formula, the LSTM network can dynamically decide which information to retain and which to forget, and this is achieved through the hidden state. Output the memory at each time step. After the LSTM network models the input data through temporal learning, it finally outputs the memory based on the LSTM hidden state. Output the final surface precipitation inversion results. Through the activation function, the LSTM hidden state... Mapped to surface precipitation inversion results ,Right now .

[0052] Regarding the radar vertical mapping error calculation method provided in this application for the urban environment, such as Figure 2 As shown, this application first uses the current radar reflectivity dataset of the target city to obtain precipitation inversion results corresponding to each altitude layer through inversion. Then, based on multiple driving factors, the altitude error characteristics of each altitude layer are solved to obtain altitude error characteristics composed of correlation index, amplitude deviation index, and random error index. Finally, based on the precipitation inversion results corresponding to each altitude layer, the altitude error characteristics corresponding to each altitude layer, and the current driving factor dataset of the target city (which records the data of various driving factors), a pre-trained surface precipitation prediction model is used for prediction processing to obtain a more accurate surface precipitation inversion result fused from multiple altitude layers.

[0053] The radar vertical mapping error calculation method provided in this application for an urban environment achieves effective integration of the advantageous information of precipitation inversion from radar reflectivity data at different altitudes under different meteorological conditions by nonlinearly and collaboratively fusing the precipitation inversion results. Simultaneously, the method introduces multi-source meteorological elements and urban underlying surface characteristics, which can influence vertical mapping error, as error-driving factors during the fusion process to correct the vertical mapping error. Compared with previous inversion results based on single-altitude radar reflectivity data for surface precipitation, this method offers higher stability and accuracy.

[0054] In this application, after step S02_04, the method may further include: Step S02_05: With the trained random forest model obtained, based on the error reduction caused by each driving factor when the decision trees in the random forest model split at the node, calculate the average contribution of each driving factor at each height layer.

[0055] In this embodiment, to improve the efficiency of solving the height error feature mapping function, after obtaining the trained random forest model, this application further calculates the average contribution of each driving factor at each height level based on the error reduction caused by each driving factor when all decision trees split at nodes in the random forest model. That is, for any driving factor, the error reduction caused by that driving factor at the same height level when all decision trees split at nodes is calculated, then the values ​​are summed and averaged to obtain the average contribution of that driving factor at that height level. Through the same implementation method, the average contribution of each driving factor at each height level can be calculated. The expression is: ,in, Indicates driving factor The average contribution at the h-th height level; T is the total number of decision trees in the random forest; This indicates that in the q-th decision tree, driven by factors... The amount of error reduction caused by participating in node splitting at the h-th height level.

[0056] Step S02_06: Among the average contributions of all driving factors at the same height level, select the target driving factor whose average contribution is greater than the contribution threshold.

[0057] In this embodiment, a contribution threshold is predefined. After obtaining the average contribution of each driving factor at each height layer, among the average contributions of all driving factors at the same height layer, it is determined which driving factors have an average contribution greater than the contribution threshold. These driving factors with an average contribution greater than the threshold are identified as the target driving factors corresponding to that height layer. Other driving factors that are not identified as target driving factors for that height layer are those with relatively small contributions to that height layer. These driving factors have a smaller impact on the accuracy of the subsequent calculation of the height error features of that height layer by the random forest model. Therefore, to improve the solution efficiency of the trained random forest model, when determining the corresponding height error features for a certain height layer, only the driving factor data of the target driving factors corresponding to that height layer are input into the trained random forest model for solution, thereby improving the solution efficiency.

[0058] For example, the driving factors include types a, b, and c, and the height layers include the first and second height layers. Calculations show that the target driving factors for the first height layer are type a and type b, meaning only the average contribution of type a and type b factors in the first height layer exceeds a certain contribution threshold. When determining the height error characteristics of the first height layer, only the driving factor data for type a and type b factors are input. Similarly, the target driving factors for the second height layer are calculated to be type b and type c, meaning only the average contribution of type b and type c factors in the second height layer exceeds a certain contribution threshold. When determining the height error characteristics of the second height layer, only the driving factor data for type b and type c factors are input.

[0059] Step S02_07: When inputting the driving factor dataset for each height layer during the application stage of the height error feature mapping function, input the driving factor dataset corresponding to the target driving factor of the height layer.

[0060] In this embodiment, after selecting each target driving factor whose average contribution is greater than the contribution threshold through step S02_06, in order to improve processing efficiency, this application inputs the driving factor dataset corresponding to the target driving factor corresponding to the height layer when inputting the driving factor dataset for each height layer during the application stage of the height error feature mapping function. That is, this application first prepares a large number of driving factor types, and then selects a portion of driving factors with larger average contributions based on the average contribution of each driving factor. Then, only the data of the driving factors with larger average contributions are input into the height error feature mapping function to calculate the height error feature. The contribution threshold is preferably 0.05, but it can also be set according to the actual application scenario, and is not specifically limited here.

[0061] For example, the driving factors include Q1 to Q10, among which the average contribution of driving factors Q2, Q6 and Q8 is greater than the contribution threshold, so these three driving factors are identified as target driving factors.

[0062] In this application, a radar reflectivity dataset for the target city is determined, including: Step S001: Convert the raw echo data of the phased array radar of the target city into the first radar reflectivity dataset of the target city.

[0063] In this embodiment, an optional implementation method for determining the radar reflectivity dataset of a target city in the current rainfall event is as follows: First, we collect the raw echo data of the phased array radar of the target city during the current rainfall event. This is the most raw data.

[0064] Then, based on the original echo data of the phased array radar and its own parameters (such as transmit power and beamwidth), the radar reflectivity data is derived by using the radar equation to obtain the first radar reflectivity dataset of the target city in the current rainfall event.

[0065] Step S002: Perform quality control processing on the first radar reflectivity dataset to obtain the second radar reflectivity dataset.

[0066] In this embodiment, the first radar reflectivity dataset obtained in step S001 is subjected to quality control processing to obtain more accurate second radar reflectivity data for the target city in the current rainfall event.

[0067] Step S003: Based on the second radar reflectivity dataset under multiple elevation angle conditions, the radar reflectivity data is uniformly mapped to a vertical coordinate system with height as the reference through volume scan reconstruction method to obtain the radar reflectivity dataset of the target city.

[0068] In this embodiment, after the quality control processing in step S002, for the second radar reflectivity dataset under multiple elevation angle conditions after quality control processing, the volume scan reconstruction method is used to uniformly map all radar reflectivity data to a vertical coordinate system based on height, so as to obtain the final radar reflectivity dataset for surface precipitation inversion of the target city.

[0069] In this application, step S002 may include: Step S002_1: Remove the radar reflectivity data in the first radar reflectivity dataset whose radar reflectivity is less than the preset clutter identification threshold to obtain a first-stage removed radar reflectivity dataset.

[0070] In this embodiment, the specific quality control process is as follows: First, radar reflectivity data with reflectivity lower than a preset clutter identification threshold are removed from the obtained first radar reflectivity dataset, resulting in a first radar reflectivity dataset. The radar reflectivity data records not only radar reflectivity but also observation time, distance, and azimuth. The preset clutter identification threshold is preferably -10 dBZ, but can be set according to the actual application scenario; no specific limitation is made here.

[0071] Step S002_2: Remove the first radar reflectivity data that meets the anomaly identification condition from the first radar reflectivity dataset to obtain the second radar reflectivity dataset. The anomaly identification condition is that the absolute value of the deviation between the first radar reflectivity data and the mean radar reflectivity in the neighborhood is greater than the preset anomaly identification threshold.

[0072] In this embodiment, for the radar reflectivity dataset obtained after the first removal in step S002_1, anomaly detection conditions are further preset. These anomaly detection conditions are that the absolute value of the deviation between the first radar reflectivity data and the mean radar reflectivity in the neighborhood is greater than a preset anomaly detection threshold. This preset anomaly detection threshold is preferably 10 dBZ, but can be set according to the actual application scenario and is not specifically limited here. Then, those first radar reflectivity data that meet the anomaly detection conditions in the first removal radar reflectivity dataset are removed. The remaining first radar reflectivity data at this point constitute the second removal radar reflectivity dataset.

[0073] Step S002_3: Based on the preset reflectivity correction value, correct each of the first radar reflectivity data in the secondary radar reflectivity dataset to obtain the second radar reflectivity dataset.

[0074] In this embodiment, after obtaining the secondary radar reflectivity dataset through step S002_2, a pre-set reflectivity correction value is added to each first radar reflectivity data in the secondary radar reflectivity dataset to obtain the corresponding corrected first radar reflectivity data. These corrected first radar reflectivity data constitute the second radar reflectivity dataset.

[0075] In this application, the method further includes: calculating the total difference field between height layers based on the precipitation inversion results corresponding to each height layer using the height layer difference field algorithm; The expression for the height inter-layer difference field algorithm is as follows:

[0076]

[0077]

[0078]

[0079]

[0080] in, This represents the total difference field between the two altitude layers of surface precipitation at time t within the target city area; This represents the difference in average precipitation intensity between two altitude levels within the target city area at time t; This represents the spatial structure difference remaining within the target city region after deducting the difference in average precipitation intensity at time t. Its value is based on the spatial structure difference within the target city region. The summation is then averaged to obtain the result. , Ω represents the average precipitation intensity at the i-th and j-th altitude layers within the target city area at time t, respectively; |Ω| represents the target city area, and |Ω| represents the number of effective grid cells within the target city area. This represents the spatial structure difference remaining at time t after deducting the difference in average rainfall intensity within the target city area (x, y); The total difference field of surface precipitation inversion at time t represents the two height layers at spatial location (x,y) within the target city area; , The values ​​represent the surface precipitation inversion results at time t for the i-th and j-th altitude layers at spatial location (x, y).

[0081] In this embodiment, the total difference between two precipitation results obtained by inverting radar reflectivity data from any two altitude layers can be calculated using the inter-altitude difference field algorithm. By using the total difference between the two precipitation results corresponding to any two altitude layers, anomalies can be assessed for all precipitation inversion results corresponding to all current altitude layers before applying the radar vertical mapping error calculation method described in this application. If anomalies are found, the precipitation inversion results corresponding to the abnormal altitude layers are removed, and then the final surface precipitation inversion result is determined using the method described in this application for the remaining altitude layers. Alternatively, if anomalies are found, the precipitation inversion result closest to the data table is used as the final precipitation inversion result.

[0082] In this embodiment, as Figure 3 As shown, this application calculates surface precipitation inversion results using three different methods for multiple rainfall events at a specific rain gauge station in a certain city. Yellow data points correspond to the surface precipitation inversion results obtained using the method described in this application, blue points correspond to the surface precipitation inversion results obtained using the BCA model, and red points correspond to the surface precipitation inversion results obtained using the Co-krining model. Figure 3It can be observed that the average deviation error of the surface precipitation inversion results for each rainfall event obtained by using the scheme of this application fluctuates around 0, which shows that it has better inversion performance.

[0083] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily necessary for the embodiments of this application.

[0084] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0085] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0087] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0088] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0089] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0090] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0091] The above provides a detailed description of a radar vertical mapping error calculation method in an urban environment. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for calculating radar vertical mapping error in an urban environment, characterized in that, The method includes: Based on the radar reflectivity dataset of the target city, the precipitation inversion results corresponding to each height layer are calculated through pre-constructed layered radar precipitation inversion models. The input of the layered radar precipitation inversion model is the radar reflectivity data of the corresponding height layer, and the output is the precipitation inversion result. Based on the driving factor dataset of the i-th altitude layer in the target city, the altitude error features of the i-th altitude layer are calculated through a pre-constructed altitude error feature mapping function. The target city's driving factor dataset, precipitation inversion results for each altitude layer, and altitude error features are input into a pre-trained surface precipitation prediction model for prediction processing to obtain the final surface precipitation inversion results for the target city. The driving factors include atmospheric dynamic driving factors, atmospheric thermal driving factors, and underlying surface driving factors. The height error feature of the i-th height layer is an N-dimensional array corresponding to the i-th height layer. The j-th element of the N-dimensional array is the vertical mapping error vector between the i-th and j-th height layers. The vertical mapping error vector includes a correlation index, an amplitude deviation index, and a random error index between the two corresponding height layers, where i and j range from 1 to N, and N is the total number of height layers. The correlation index is an indicator of the consistency of the precipitation inversion results of the two corresponding height layers in terms of spatial distribution structure. The amplitude deviation index is an indicator of the systematic difference in amplitude of the precipitation inversion results of the two corresponding height layers. The random error index is a random error index of the precipitation inversion results of the two corresponding height layers at the local scale. The construction of the height error feature mapping function includes: Construct an initial random forest model, the input of which is the height layer and the corresponding driving factor dataset, and the output is the height error feature corresponding to the height layer; The driving factor dataset of the i-th height layer of historical rainfall events is used as input and imported into the initial random forest model to obtain the predicted height error features of the i-th height layer; The residuals between the predicted height error features and the standard height error features at the same height level are calculated. The initial random forest model is iteratively optimized with the goal of minimizing the sum of all residuals until the model accuracy reaches a preset threshold, thus obtaining a trained random forest model. The standard height error features at each height level are calculated based on the radar reflectivity dataset of the historical rainfall events. The trained random forest model is determined to be a height error feature mapping function.

2. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, Each layer of radar precipitation inversion model is pre-constructed, corresponding to a different altitude level. The expression for each layer of radar precipitation inversion model is as follows: in, Represents the radar reflectivity data of the spatial position (x, y) in the plane at time t in the h-th spatial layer; α and β represent evolution parameters; The precipitation inversion result is obtained by inverting the radar reflectivity data of the plane spatial location (x,y) at time t in the h-th spatial layer.

3. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, Based on the radar reflectivity dataset of the historical rainfall events, the standard height error characteristics for each altitude layer are calculated, including: Based on the radar reflectivity dataset of the historical rainfall events, the historical precipitation inversion results corresponding to each height layer are calculated by using the pre-constructed layered radar precipitation inversion models. Based on the historical precipitation inversion results, the vertical mapping error vector between the i-th height layer and the j-th height layer is calculated; Based on all the calculated vertical mapping error vectors, standard height error features corresponding to each height layer are generated. The expression for calculating the correlation index in the vertical mapping error vector is as follows: ,in, is the correlation index between the i-th and j-th altitude layers; n is the total number of valid spatial sampling points within the target city area; , These represent the precipitation inversion values ​​at the k-th spatial location for the i-th and j-th altitude layers, respectively. , represent the average precipitation inversion values ​​of the i-th and j-th altitude layers within the target city area, respectively; The expression for calculating the magnitude deviation index in the vertical mapping error vector is as follows: ,in, This is the magnitude deviation index between the i-th and j-th height layers; The expression for calculating the random error index in the vertical mapping error vector is as follows: ,in, This is the random error index between the i-th and j-th altitude layers.

4. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, The atmospheric dynamic driving factors include wind speed and wind direction factors; the atmospheric thermal driving factors include inlet pressure, air temperature and relative humidity factors; the underlying surface driving factors include high vegetation leaf area index, low vegetation leaf area index and impermeable surface ratio.

5. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, The method further includes: Given a trained random forest model, the average contribution of each driving factor at each height level is calculated based on the error reduction caused by each driving factor when all decision trees in the random forest model split at nodes. Among the average contributions of all driving factors at the same height level, select the target driving factor whose average contribution is greater than the contribution threshold. When inputting the driving factor dataset for each height layer during the application phase of the height error feature mapping function, the driving factor dataset corresponding to the target driving factor for the height layer is input.

6. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, Determine the radar reflectivity dataset for the target city, including: The raw echo data of the phased array radar of the target city is collected and transformed into the first radar reflectivity dataset of the target city. The first radar reflectivity dataset is subjected to quality control processing to obtain the second radar reflectivity dataset. Based on the second radar reflectivity dataset under multiple elevation angle conditions, the radar reflectivity data is uniformly mapped to a vertical coordinate system with height as the reference through volume scan reconstruction method to obtain the radar reflectivity dataset of the target city.

7. The method for calculating radar vertical mapping error in an urban environment according to claim 6, characterized in that, The first radar reflectivity dataset is subjected to quality control processing to obtain a second radar reflectivity dataset, including: The radar reflectivity data in the first radar reflectivity dataset with radar reflectivity less than the preset clutter identification threshold are removed to obtain a first-stage removed radar reflectivity dataset. The first radar reflectivity data that meets the anomaly identification condition in the first radar reflectivity dataset is removed to obtain the second radar reflectivity dataset. The anomaly identification condition is that the absolute value of the deviation between the first radar reflectivity data and the mean radar reflectivity in the neighborhood is greater than the preset anomaly identification threshold. Based on the preset reflectivity correction value, each first radar reflectivity data in the secondary radar reflectivity dataset is corrected to obtain the second radar reflectivity dataset.

8. The method for calculating radar vertical mapping error in an urban environment according to claim 1, characterized in that, The method further includes: Based on the precipitation inversion results corresponding to each altitude layer, the total difference field between altitude layers is calculated using the altitude-inter-altitude difference field algorithm. The expression for the height inter-layer difference field algorithm is as follows: in, This represents the total difference field between the two altitude layers of surface precipitation at time t within the target city area; This represents the difference in average precipitation intensity between two altitude levels within the target city area at time t; This represents the spatial structure difference remaining within the target city region after deducting the difference in average precipitation intensity at time t. Its value is based on the spatial structure difference within the target city region. The summation is then averaged to obtain the result. , Ω represents the average precipitation intensity at the i-th and j-th altitude layers within the target city area at time t, respectively; |Ω| represents the target city area, and |Ω| represents the number of effective grid cells within the target city area. This represents the spatial structure difference remaining at time t after deducting the difference in average rainfall intensity within the target city area (x, y); The total difference field of surface precipitation inversion at time t represents the two height layers at spatial location (x,y) within the target city area; , The values ​​represent the surface precipitation inversion results at time t for the i-th and j-th altitude layers at spatial location (x, y).

Images