A method for 0–6 h precipitation forecasting that integrates numerical models and radar monitoring

Abstract

This invention discloses a 0-6 hour precipitation forecasting method that integrates numerical model and radar monitoring. It utilizes radar extrapolation to estimate precipitation, which is then used to correct the phase of the predicted precipitation field from the numerical model. The intensity of the predicted precipitation field is then corrected using a Weiper distribution function. The corrected predicted precipitation field and the precipitation field estimated by the radar extrapolation method are then fused and analyzed to generate a 0-6 hour precipitation forecast. The advantages of this invention lie in its use of blending technology to filter out scattered precipitation areas and remove false precipitation zones from the numerical model. By comprehensively considering precipitation information from both radar extrapolation and numerical model forecasts, it significantly improves the forecast quality of precipitation area, extent, and intensity in numerical model forecasts.

Description

Technical Field

[0001] This invention relates to the field of precipitation forecasting technology, and in particular to a 0-6 hour precipitation forecasting method that integrates numerical models and radar monitoring. Background Technology

[0002] The foundation of heavy precipitation forecasting is high-quality quantitative precipitation forecasting (QPF). Currently, the mainstream methods for 0–12 hour short-term quantitative precipitation forecasting include 0–2 hour nowcasting based on observational data and 3–12 hour mesoscale numerical model forecasting based on rapid assimilation of observational data. These two methods differ in forecast lead time and accuracy. 0–2 hour nowcasting mainly relies on the identification of convective storms and severe convective weather to provide 0–2 hour nowcasting. 3–12 hour mesoscale numerical model forecasting mainly relies on high-resolution numerical prediction models, but its accuracy for 0–3 hour nowcasting is low and it cannot obtain forecast products in a timely manner. To improve the accuracy of 0–6 hour nowcasting, a common technique is to fuse 0–2 hour nowcasting with 3–12 hour mesoscale numerical model forecasts. The key to fusion lies in how to determine the optimal weighting coefficients for 0–2h short-term nowcast extrapolation forecasts and 3–12h mesoscale numerical model forecasts, so as to fully take into account the advantages of 0–2h short-term nowcast extrapolation forecasts and 3–12h mesoscale numerical model forecasts in terms of their respective forecast lead time and accuracy, thereby achieving the highest relative accuracy of short-term nowcast quantitative precipitation forecasts throughout the entire 0–6h forecast lead time.

[0003] However, most current short-term and nowcasting systems use TREC (Chinese definition: echo cross-correlation analysis technology) for 0-2 hour short-term and nowcasting extrapolation and phase correlation (English full name: phase correlation) technology to correct 3-12 hour mesoscale numerical model forecasts. They also use phase correlation technology to adjust the bias of the mesoscale numerical model system based on the observed phase and intensity of precipitation systems, thereby improving the accuracy of short-term and nowcasting.

[0004] In recent years, my country has also begun research on 0–6 hour short-term quantitative precipitation forecasting. For example, some researchers have proposed a precipitation probability fusion forecasting method based on dynamic weights. This method can simultaneously consider both precipitation location and precipitation amount, and dynamically allocate the weights of 0–2 hour short-term extrapolation forecasts and 3–12 hour mesoscale numerical model forecasts at different forecast lead times. This ensures that the fused forecasts at each lead time exhibit technical scores similar to or even higher than those of radar extrapolation or numerical models.

[0005] This application also focuses on 0-6 h short-term now-quantitative precipitation forecasting, and proposes a method for 0-6 h short-term now-quantitative precipitation forecasting that integrates numerical models and radar monitoring. Summary of the Invention

[0006] The purpose of this invention is to provide a method for 0-6 h precipitation forecasting that integrates numerical models and radar monitoring.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The 0-6 h precipitation forecasting method integrating numerical models and radar monitoring described in this invention includes the following steps:

[0009] S1, the precipitation is estimated using radar extrapolation forecasting method, and the estimated precipitation is used to correct the phase of the forecast precipitation field of numerical model forecasting method;

[0010] S2, The intensity of the predicted precipitation field in the numerical model prediction method is corrected by the Weiber distribution function;

[0011] S3 integrates the corrected forecast precipitation field with the precipitation field estimated by the radar extrapolation forecast method to perform a 0-6 hour precipitation forecast.

[0012] Furthermore, the method for forecasting the phase of the precipitation field in the modified numerical model forecasting method includes the fast Fourier transform method, the target identification analysis method, the target feature description and matching method, and the combination of fast Fourier transform and objective diagnostic evaluation.

[0013] Furthermore, the Weber distribution function is obtained using the least squares method.

[0014] Furthermore, the fusion analysis of the modified forecast precipitation field and the precipitation field estimated by the radar extrapolation forecast method uses a hyperbolic tangent function to dynamically adjust the weights of the numerical model forecast method predictions and the radar extrapolation forecast method predictions within different forecast lead times.

[0015] Furthermore, the Fast Fourier Transform method specifically includes: assuming the predicted precipitation field of the numerical model prediction method at time T is... The estimated precipitation field of the time-based radar extrapolation forecast method is: Using Discrete Fourier Transform to and Transform to the frequency domain, assuming Only by By simple translation and ignoring the change in precipitation field intensity, a two-dimensional pulse function is obtained through inverse transformation. The peak position of the phase correlation coefficient in the two-dimensional pulse function is then determined to identify the translation parameters of the predicted precipitation field phase in the modified numerical model forecasting method.

[0016] Furthermore, the target identification and analysis method specifically includes: performing convolution operations on the precipitation areas of the predicted precipitation field of the numerical model forecasting method and the estimated precipitation field of the radar extrapolation forecasting method, respectively; using a precipitation threshold T, binarizing the precipitation areas and outlining the boundaries of the precipitation areas; and correcting the phase of the predicted precipitation field of the numerical model forecasting method by comparing the boundaries of the precipitation areas.

[0017] Furthermore, the target feature description and matching method specifically includes: calculating the attributes and matching degree of the precipitation area of ​​the forecast precipitation field of the numerical model prediction method and the estimated precipitation field of the radar extrapolation prediction method, respectively; when the matching degree score of the precipitation area is greater than 0.7, the precipitation area of ​​the forecast precipitation field of the numerical model prediction method and the estimated precipitation field of the radar extrapolation prediction method are considered to be matched; by calculating the distance of the matched precipitation area, the total displacement of the phase of the forecast precipitation field of the corrected numerical model prediction method is obtained.

[0018] Furthermore, the method combining fast Fourier transform and objective diagnostic evaluation specifically includes: using the objective diagnostic evaluation method to delineate the boundaries of the predicted precipitation field of the numerical model forecasting method and the estimated precipitation field of the radar extrapolation forecasting method within the same target area, and then using the fast Fourier transform method to determine the phase shift parameters of the predicted precipitation field of the modified numerical model forecasting method.

[0019] Furthermore, the attributes include the precipitation center, i.e., the location, area, and intensity of the target precipitation.

[0020] The advantages of this invention lie in its use of blending technology to filter out scattered precipitation areas and eliminate false precipitation areas from numerical models. By comprehensively considering precipitation information from radar extrapolation and numerical model forecasts, it significantly improves the forecast quality of precipitation area, extent, and intensity in numerical model forecast products. This invention corrects the phase and intensity of numerical forecast products for precipitation systems up to 6 hours in advance. In the early stages (2-4 hours lead time), the forecast is closer to the short-term nowcast conclusion, with a larger correction to the numerical model forecast. Later, as the short-term nowcast information weakens, the fusion result becomes more biased towards the corrected numerical model forecast result. Analysis of the results shows that the maximum offset of the precipitation area can be controlled between 0.03° and 0.05° (approximately 3km to 5km), and the overall hit rate is improved by 0.1-0.2% compared to numerical model products. In other words, the fusion forecasting algorithm of this invention is generally better than the performance of simple numerical models or radar nowcasting for hourly quantitative precipitation forecasting. It improves the accuracy of precipitation area forecasting to a certain extent and also enhances the accuracy of precipitation intensity forecasting. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method described in this invention. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] The 0-6 h precipitation forecasting method merging numerical models and radar monitoring described in this invention requires a two-step correction of the numerical model forecast results before fusing the numerical model and radar monitoring forecasts. This correction includes adjustments to the location and intensity of the precipitation area predicted by the numerical model. For example... Figure 1 As shown, the specific steps include:

[0024] S1 uses radar extrapolation forecasting to estimate precipitation. This estimated precipitation is then used to correct the phase of the precipitation field predicted by numerical model forecasting methods. Specifically:

[0025] The correction methods include Fast Fourier Transform (FFT), objective diagnostic assessment, target feature description and matching, and a combination of FFT and objective diagnostic assessment.

[0026] The Fast Fourier Transform method specifically includes: assuming the predicted precipitation field at time T is... The estimated precipitation field of the time-based radar extrapolation forecast method is: Using Discrete Fourier Transform to and Transforming to the frequency domain, the formula is as follows:

[0027]

[0028]

[0029] in, The grid coordinates of a certain location after Fourier transforming the original precipitation field in the numerical model forecast method. These are the grid coordinates of a certain location in the raw precipitation field of the numerical model forecasting method. Take a natural number between 0 and N-1. N is the total number of grid point coordinates. j is the imaginary unit of the formula. Assume... Only by After a simple translation, ignoring the transformation of precipitation field intensity, we can obtain the following based on the properties of the Fourier transform:

[0030]

[0031] In the formula and They are respectively and Fourier transform, and These are the required correction parameters for the precipitation field in the numerical model. and The cross-power spectrum is:

[0032]

[0033] In the above formula, yes The complex conjugate of . (The following is a conjugate of .) Inverse transform yields the two-dimensional impulse function. By determining the peak position of the phase correlation coefficient of the two-dimensional pulse function, the correction parameters are finally determined. and By adjusting the parameters and The numerical model precipitation field is positionally corrected to obtain the corrected numerical model precipitation field.

[0034] The target identification and analysis method specifically includes: first, identifying the target precipitation area for diagnostic assessment; then, describing the characteristics of the target precipitation area, such as the intensity and shape of the radar-estimated precipitation field and the model-predicted precipitation field. Specifically, this process involves: first, performing a convolution operation on the predicted precipitation field from the numerical model prediction method, using the following convolution function:

[0035]

[0036] In the formula, This is the original precipitation field predicted by the numerical model. It is a filter function, and the variables are... These are the grid coordinates of a certain location in the raw precipitation field of the numerical model forecasting method. The grid coordinates of a certain location after Fourier transforming the original precipitation field in the numerical model forecasting method.

[0037] Filtering function It is a circular filter determined by the radius R. The purpose of convolution is to smooth the forecast precipitation field of the numerical model, making it more continuous and filtering out some small-scale systems that are not predictable. Then, a precipitation threshold T is used to binarize the forecast precipitation field. After binarization, sporadic and weak precipitation in the forecast precipitation field will be filtered out, making the target area for diagnosis and assessment more prominent. The boundary of the target area can be delineated through the binarized forecast precipitation field.

[0038] Then, the same method is used to perform convolution operations on the radar-estimated precipitation field to delineate the boundary of the target area. By objectively evaluating the target area boundaries delineated by the radar estimation method and the model prediction method, the phase of the predicted precipitation field of the numerical model prediction method is corrected.

[0039] The target feature description and matching method specifically includes: after identifying the precipitation areas in the radar-estimated precipitation field and the numerical model-predicted precipitation field, attribute calculations are performed on the precipitation areas in both fields. Attributes include the precipitation center, i.e., the target precipitation location, area, and intensity. Then, targets in the radar-estimated precipitation field and the numerical model-predicted precipitation field are matched. The feature quantities used in the matching include the centroid distance deviation between the two targets, the shortest distance between the target boundaries, the tilt angle deviation, the ratio of overlapping target areas, and the target area ratio, with each feature quantity having a different weight. After the target matching in the radar-estimated precipitation field and the numerical model-predicted precipitation field is completed through fuzzy logic calculation, the value function is calculated. , the formula is as follows:

[0040]

[0041] in, This represents the confidence distribution of the i-th feature, reflecting the credibility of the i-th feature. It is the membership function of the i-th feature of the j-th target. is the weight of the i-th feature. M is the total number of features. The larger the value, the greater the similarity between the targets. A score greater than 0.7 indicates a successful pairing. The distance between successfully paired targets is calculated. This allows us to obtain the total displacement of the precipitation field. .

[0042] The calculation formula is as follows:

[0043]

[0044] Where P is the total number of successful pairings.

[0045] The combination of Fast Fourier Transform and objective diagnostic assessment method specifically includes: using objective diagnostic assessment methods to delineate the boundaries of the numerical model-predicted precipitation field and the radar extrapolation-estimated precipitation field within the same target area, and then using the Fast Fourier Transform method to determine the phase shift parameters of the predicted precipitation field of the modified numerical model prediction method.

[0046] S2, after phase correction of the numerical model's predicted precipitation field, further corrections are made to the intensity of the predicted precipitation field. The numerical model's predicted precipitation field is adjusted using the Weiper distribution function of the radar extrapolation forecast. The Weiper distribution function is the most widely used probability function for reliability evaluation, exhibiting strong fitting accuracy and adaptability for the probabilistic characteristics of meteorological random variables such as precipitation and visibility. Its formula is:

[0047]

[0048] in, A value greater than 0 indicates a shape parameter. A value greater than 0 indicates a scale parameter. For position parameters, in When this happens, solving for the Weiber distribution function becomes a two-parameter problem. and The Weiber distribution function is used. This invention employs the least squares estimation method to obtain relevant parameters. Assuming that both the radar extrapolation forecast and the numerical model forecast precipitation field follow a Weiber distribution, the numerical model forecast precipitation field can be adjusted based on the Weiber distribution function obtained from the radar extrapolation forecast. The specific relationship between the two is as follows:

[0049]

[0050] in, To adjust the precipitation field forecast by the numerical model, For numerical model forecasting of precipitation fields, The Weiber distribution function for radar extrapolation forecasting of precipitation fields; This is the Weiber distribution function of the precipitation field predicted by the numerical model.

[0051] S3 involves fusing the corrected forecast precipitation field with the precipitation field estimated by the radar extrapolation method to generate a 0-6 hour precipitation forecast. The fusion analysis of the corrected forecast precipitation field and the precipitation field estimated by the radar extrapolation method uses a hyperbolic tangent function to dynamically adjust the weights of the numerical model predictions and the radar extrapolation predictions within different forecast lead times. In shorter forecast lead times, the precipitation estimated by the radar extrapolation method has a larger weight; as the forecast lead time increases, the weight of the numerical model predictions increases. The specific formula is as follows:

[0052]

[0053] In the formula, , This represents the fused rainfall forecast at time t. This represents the radar extrapolation forecast method for estimating precipitation at time t. This represents the rainfall forecast using a numerical model at time t. The weight coefficients of the pattern.

[0054] The weights are calculated using the empirical equation of the hyperbolic tangent function:

[0055]

[0056] in and These respectively represent numerical model forecasting methods or radar extrapolation forecasting methods in and The weights at each time point are adjusted to ensure a smooth change in the weight curve. The value is set to 1. In practical applications, it was found that the radar extrapolation forecast of precipitation has a larger weight in the first three hours of the forecast, while the weight of the numerical model forecast of precipitation increases as time increases.

Claims

1. A method for 0–6 h precipitation forecasting that integrates numerical models and radar monitoring, characterized in that: Includes the following steps: S1, the precipitation is estimated using radar extrapolation forecasting method, and the estimated precipitation is used to correct the phase of the forecast precipitation field of numerical model forecasting method; The methods for forecasting precipitation field phases in the modified numerical model forecasting method include the Fast Fourier Transform method, target identification and analysis method, target feature description and matching method, and a combination of Fast Fourier Transform and objective diagnostic assessment method. The target identification and analysis method specifically includes: performing convolution operations on the precipitation areas of the predicted precipitation field of the numerical model forecasting method and the estimated precipitation field of the radar extrapolation forecasting method, respectively; using a precipitation threshold T, binarizing the precipitation areas and outlining the boundaries of the precipitation areas; and correcting the phase of the predicted precipitation field of the numerical model forecasting method by comparing the boundaries of the precipitation areas. The combined method of Fast Fourier Transform and objective diagnostic evaluation specifically includes: using the objective diagnostic evaluation method to delineate the boundaries of the predicted precipitation field of the numerical model forecasting method and the estimated precipitation field of the radar extrapolation forecasting method within the same target area, assuming that the precipitation areas of the predicted precipitation field of the numerical model forecasting method and the estimated precipitation field of the radar extrapolation forecasting method satisfy the matching; and obtaining the total displacement of the phase of the predicted precipitation field of the corrected numerical model forecasting method by calculating the distance of the matching precipitation area. S2, the intensity of the predicted precipitation field is corrected by the Weiber distribution function; the parameter values ​​of the Weiber distribution function are obtained by the least squares method; S3 uses a hyperbolic tangent function to dynamically adjust the weights of the predicted values ​​from the numerical model forecasting method and the radar extrapolation forecasting method within different forecast lead times. It then integrates and analyzes the corrected forecast precipitation field and the precipitation field estimated by the radar extrapolation forecasting method to perform 0-6 h precipitation forecasts.

2. The 0-6 h precipitation forecasting method integrating numerical model and radar monitoring according to claim 1, characterized in that: The Fast Fourier Transform method specifically includes: assuming the predicted precipitation field at time T is... The estimated precipitation field of the time-based radar extrapolation forecast method is: Using Discrete Fourier Transform to and Transform to the frequency domain, assuming Only by By simple translation and ignoring the change in precipitation field intensity, a two-dimensional pulse function is obtained through inverse transformation. The peak position of the phase correlation coefficient in the two-dimensional pulse function is then determined to identify the translation parameters of the predicted precipitation field phase in the modified numerical model forecasting method.

3. The 0-6 h precipitation forecasting method integrating numerical model and radar monitoring according to claim 1, characterized in that: The target feature description and matching method specifically includes: calculating the attributes and matching degree of the precipitation area of ​​the predicted precipitation field of the numerical model forecast method and the estimated precipitation field of the radar extrapolation forecast method respectively; when the matching degree score of the precipitation area is greater than 0.7, the total displacement of the phase of the predicted precipitation field of the corrected numerical model forecast method is obtained by calculating the distance of the precipitation area.

4. The 0-6 h precipitation forecasting method integrating numerical model and radar monitoring according to claim 3, characterized in that: The attributes include the precipitation center, i.e., the location, area, and intensity of the target precipitation.

Images