A dual-resolution model typhoon forecast post-processing method and system

Abstract

The present invention discloses a dual-resolution model typhoon forecast post-processing method and system. The post-processing method corrects the typhoon path forecast in the global-scale ensemble forecast to obtain a corrected typhoon path forecast and precipitation forecast, and performs probability conversion on the mesoscale deterministic forecast. Finally, the corrected typhoon precipitation ensemble forecast and the mesoscale probabilistic precipitation forecast are integrated, giving full play to the advantages of the global-scale ensemble forecast and the mesoscale deterministic forecast to obtain a precipitation forecast result with higher accuracy. The post-processing method of the present invention first corrects the global-scale ensemble forecast data, screens out members with better forecast quality in the global-scale ensemble forecast, and forms a corrected typhoon path and precipitation ensemble forecast based on the typhoon forecasts of the better-quality members, thereby improving the accuracy of the ensemble forecast and further improving the accuracy of the forecast post-processing results.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological forecasting, and in particular relates to a dual-resolution model typhoon forecast post-processing method and system. Background Art

[0002] Currently, the research and application of ensemble forecasts to typhoons focuses on two main areas: the establishment of typhoon ensemble forecast systems; and the application of ensemble forecast products in post-processing and correction (i.e., ensemble forecast interpretation). Overall, research on typhoon ensemble forecast post-processing and correction remains relatively limited, primarily focusing on corrections to global-scale model ensemble forecasts.

[0003] The heavy rainfall caused by typhoons is not only related to the intensity, scale, and structure of the typhoon itself, but also to the multi-scale interactions of the environmental field and the characteristics of the underlying surface. In current typhoon forecasting research, global-scale ensemble forecasts simulate large-scale environmental fields better than other scales / models, and can accurately simulate typhoon tracks. However, due to their lower resolution, global-scale ensemble forecasts cannot accurately simulate the internal structure of the typhoon vortex and the influence of topography on precipitation. This can result in systematic underestimation due to insufficient model resolution or systematic overestimation due to strong topographic effects. Meanwhile, mesoscale models can more accurately simulate the mesoscale characteristics of typhoon precipitation and the influence of topography on precipitation, but they still have significant errors in the prediction of typhoon tracks and precipitation areas. Therefore, some researchers have developed ensemble forecasts of typhoon precipitation by fusing ensemble forecasts of different resolutions. Specifically, low-resolution global-scale ensemble forecasts are used to estimate the optimal typhoon track, thereby correcting the spatial structure of precipitation distribution, and high-resolution mesoscale ensemble forecasts are used to adjust heavy rain amount forecasts. However, most current fusion methods directly fuse ensemble forecasts, which inherently have certain inaccuracies. Directly fusing ensemble forecasts can be affected by the poor quality of the component forecasts within the ensemble, reducing the accuracy of the final fused forecast. Therefore, one study combined and corrected low-resolution (36 km) and high-resolution (4 km) ensemble forecasts to forecast the precipitation from Typhoon Morakot in Taiwan Province, my country. The results showed that the combined ensemble precipitation forecast reduced the errors of both the high-resolution and low-resolution ensemble forecasts. However, the direct fusion of the two resolutions also improved the accuracy of the combined ensemble precipitation forecast to a certain extent. At the same time, this combined correction method for high-resolution ensemble forecasts presents difficulties in acquiring data sources; for example, high-resolution ensemble forecasts are difficult to obtain. For provincial and municipal forecast operations, various limitations hinder the accuracy of local high-resolution ensemble forecasts, leading to the increased use of existing high-resolution deterministic forecasts and related correction products. Summary of the Invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a dual-resolution model typhoon forecast post-processing method and system.

[0005] The present invention adopts the following technical solutions:

[0006] A dual-resolution model typhoon forecast post-processing method, wherein the dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched; the post-processing method comprises:

[0007] Combining historical typhoon data in the target area, the typhoon track forecast in the global-scale ensemble forecast is corrected in terms of forecast quality;

[0008] Perform probability conversion on mesoscale deterministic forecasts to form mesoscale probabilistic forecasts;

[0009] The revised typhoon track forecast is integrated with the mesoscale probability forecast.

[0010] Furthermore, the specific method for correcting the forecast quality of the global-scale ensemble forecast by combining historical typhoon data of the target area includes:

[0011] Based on the historical typhoon data of the target area, the members of the global-scale ensemble forecast are screened in terms of forecast quality. Based on the screened members, a revised typhoon path forecast and corresponding typhoon precipitation forecast are formed.

[0012] Furthermore, based on the historical typhoon data of the target area, the members of the global-scale ensemble forecast are screened for their forecast quality, and based on the screened members, a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed using the following method:

[0013] 1) Obtain the real-time location and subjective forecast of the typhoon impact process from the historical typhoon data of the target area;

[0014] 2) Based on the global-scale ensemble forecast data, the distance errors between the path forecast of each member in the global-scale ensemble forecast and the real-time position and subjective forecast during the typhoon impact process are calculated, and the N members with the smallest errors are selected;

[0015] 3) The arithmetic mean of the typhoon path forecasts of the screened N members is used as the revised typhoon path forecast and the corresponding typhoon precipitation forecast.

[0016] Furthermore, the probability conversion of the mesoscale deterministic forecast is performed using the following method:

[0017] The ROC curve areas of the mesoscale deterministic forecast at different scales are calculated, and the scale corresponding to the ROC curve with the largest area is selected as the neighborhood scale. The mesoscale deterministic forecast is then converted into a probability according to the selected neighborhood scale.

[0018] Furthermore, the revised typhoon path forecast and the mesoscale probability forecast after probability conversion are fused using a frequency matching method.

[0019] Furthermore, the frequency matching fusion method includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast as the typhoon precipitation area forecast, using the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion.

[0020] Furthermore, the frequency matching fusion method includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast and the mesoscale deterministic forecast as the typhoon precipitation area forecast, and using the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion.

[0021] A dual-resolution model typhoon forecast post-processing system, wherein the dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched; the post-processing system includes a correction module, a probability conversion module, and a fusion module;

[0022] The correction module is used to correct the typhoon path forecast in the global-scale ensemble forecast in relation to the forecast quality in combination with the historical typhoon data of the target area;

[0023] The probability conversion module is used to perform probability conversion on the mesoscale deterministic forecast to form a mesoscale probabilistic forecast;

[0024] The fusion module is used to fuse the revised typhoon path forecast and the mesoscale probability forecast.

[0025] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute any of the above methods.

[0026] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention provides a dual-resolution model typhoon forecast post-processing method and system. The post-processing method corrects the typhoon path forecast in the global-scale ensemble forecast to obtain a corrected typhoon path forecast and precipitation forecast, and performs probability conversion on the mesoscale deterministic forecast. Finally, the corrected typhoon precipitation ensemble forecast and the mesoscale probabilistic precipitation forecast are integrated, giving full play to the advantages of the global-scale ensemble forecast and the mesoscale deterministic forecast to obtain a precipitation forecast result with higher accuracy.

[0029] The post-processing method of the present invention first corrects the global-scale ensemble forecast data, screens out members with better forecast quality in the global-scale ensemble forecast, and forms a corrected typhoon path and precipitation ensemble forecast based on the typhoon forecasts of the better-quality members, thereby improving the accuracy of the ensemble forecast and further improving the accuracy of the forecast post-processing results.

[0030] The post-processing method of the present invention does not directly use the mesoscale ensemble forecast data, but instead performs a probability conversion on the mesoscale deterministic forecast to form a mesoscale probabilistic forecast. Therefore, it can reduce the requirements for computer software and hardware, reduce the difficulty of processing for technical personnel, and make the post-processing method of the present invention fast and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Flowchart of the correction method for typhoon track and precipitation forecasts.

[0032] Figure 2(a) shows the path of Typhoon Lekima every three hours from 14:00 on August 4 to 08:00 on August 13;

[0033] Figure 2(b) shows the cumulative precipitation map of Typhoon Lekima in East China from 08:00 on August 9 to 08:00 on August 12;

[0034] Figure 3(a) shows the spatial distribution of the cumulative precipitation of the global-scale ensemble forecast (EC-eps mean) from 08:00 on August 9 to 08:00 on August 12;

[0035] Figure 3(b) shows the spatial distribution of the mesoscale deterministic forecast (WARMS) cumulative precipitation from 08:00 on August 9 to 08:00 on August 12;

[0036] Figure 4 This is a comparison chart of the time-sensitive path error between the ensemble forecast generated using the subjective and objective member selection method and the original ensemble forecast at 6-hour intervals for Typhoon Lekima.

[0037] Figure 5(a) shows the ensemble forecast for Typhoon Lekima generated using the subjective and objective member selection method and the original ensemble forecast at 6-hour intervals, corresponding to the precipitation RPS score of 0.1 mm / 6 h precipitation.

[0038] Figure 5(b) shows the ensemble forecast for Typhoon Lekima generated using the subjective and objective member selection method and the original ensemble forecast at 6-hour intervals, corresponding to the precipitation RPS score of 5 mm / 6 h precipitation.

[0039] Figure 5(c) shows the ensemble forecast for Typhoon Lekima generated using the subjective and objective member selection method and the original ensemble forecast at 6-hour intervals, corresponding to the precipitation RPS score of 10 mm / 6 h precipitation.

[0040] Figure 5(d) shows the ensemble forecast for Typhoon Lekima generated using the subjective and objective member selection method and the original ensemble forecast at 6-hour intervals, corresponding to the precipitation RPS score of 25 mm / 6 h.

[0041] Figure 6 This is the spatial distribution of the difference in precipitation RPS scores between the original and revised ensemble forecasts for Typhoon Lekima.

[0042] Figure 7(a) shows the comparison of S1, S2, EC-eps average and WARMS ETS scores corresponding to 0.1 mm precipitation during the precipitation process of Typhoon Lekima;

[0043] Figure 7(b) shows the comparison of S1, S2, EC-eps average and WARMS ETS scores corresponding to 5 mm precipitation during the precipitation process of Typhoon Lekima;

[0044] Figure 7(c) shows the comparison of S1, S2, EC-eps average and WARMS ETS scores corresponding to 10 mm precipitation during the precipitation process of Typhoon Lekima;

[0045] Figure 7(d) shows the comparison of S1, S2, EC-eps average and WARMS ETS scores corresponding to 25 mm precipitation during the precipitation process of Typhoon Lekima. DETAILED DESCRIPTION

[0046] The present invention will be further described below with reference to specific embodiments and corresponding drawings.

[0047] Example 1:

[0048] The present invention provides a dual-resolution model typhoon forecast post-processing method, wherein the dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched. The global-scale ensemble forecast uses the ECMWF ensemble forecast including 51 members; Figure 1 As shown, the post-processing method includes:

[0049] The typhoon track forecast in the global-scale ensemble forecast is corrected in terms of forecast quality by combining historical typhoon data in the target area. That is, the typhoon track correction method is used to screen out members of the global-scale ensemble forecast with better quality and closer to the actual situation. Then, based on these members (the corresponding forecasts), a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed to improve the accuracy of the ensemble forecast.

[0050] Perform probability conversion on mesoscale deterministic forecasts to form mesoscale probabilistic forecasts;

[0051] The revised typhoon track forecast is integrated with the mesoscale probability forecast.

[0052] Example 2:

[0053] This example is further designed based on the first embodiment in that the specific method for correcting the forecast quality of the global-scale ensemble forecast in combination with the historical typhoon data of the target area includes:

[0054] Based on the historical typhoon data of the target area, the members of the global-scale ensemble forecast are screened in terms of forecast quality. Based on the forecasts corresponding to the screened members, the revised typhoon path forecast and corresponding typhoon precipitation forecast are formed.

[0055] Example 3:

[0056] This example is further designed based on the first embodiment in that, based on the historical typhoon data of the target area, members of the global-scale ensemble forecast are screened for forecast quality. Based on the forecasts corresponding to the screened members, a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed using the following method:

[0057] 1) Obtain the real-time location and subjective forecast of the typhoon impact process from the historical typhoon data of the target area;

[0058] 2) Based on the global-scale ensemble forecast data, the distance errors between the path forecast of each member in the global-scale ensemble forecast and the real-time position and subjective forecast during the typhoon impact process are calculated, and the N members with the smallest errors are selected;

[0059] 3) The arithmetic mean of the typhoon path forecasts of the screened N members is used as the revised typhoon path forecast and the corresponding typhoon precipitation forecast.

[0060] The following example specifically illustrates the screening and correction of forecast quality. This example collects individual typhoons that made landfall along China's southeastern coast over the past 10 years as historical typhoon data, and uses the Central Meteorological Observatory's forecast as the subjective forecast.

[0061] Based on historical typhoon data, the global-scale ensemble forecast ECMWF-eps is used to calculate the path errors of different member numbers through the following subjective and objective fusion member selection method, and the N members with the smallest errors are selected. The subjective and objective fusion member selection method includes:

[0062] The corrected path F is obtained by taking the arithmetic average AVE of the N paths with the smallest error among the typhoon's real-time position (subjective forecast, the real-time position of the Central Meteorological Observatory) and the subjective forecast of the Central Meteorological Observatory for the next 72 hours, the typhoon path L0, the latest position of the ECMWF ensemble forecast, and the typhoon path LE forecast for the next 72 hours. track , as shown below:

[0063]

[0064] Example 4:

[0065] This example is further designed based on the first embodiment in that the probability conversion of the mesoscale deterministic forecast is performed using the following method:

[0066] The ROC curve areas of the mesoscale deterministic forecast at different scales are calculated, and the scale corresponding to the ROC curve with the largest area is selected as the neighborhood scale. The mesoscale deterministic forecast is then converted into a probability according to the selected neighborhood scale.

[0067] The following example illustrates the specific steps of the neighborhood method. The calculation scope in this example is East China, and the calculation period is the first six hours of the precipitation forecast. The ROC area for the mesoscale deterministic forecast model WARMS is calculated for each scale range within each time period.

[0068] The function of the hit rate (HR) versus the false alarm rate (FAR) is obtained as the ROC curve. The ROC curve close to the upper left corner is usually accompanied by a larger ROC area. The neighborhood scale corresponding to the largest ROC curve is selected.

[0069]

[0070]

[0071] Where a and b represent the number of hits and misses, FA represents the number of false alarms, and CN represents the number of correct negatives.

[0072] The mesoscale deterministic forecast WARMS is converted into the probabilistic forecast WARMS-eps according to the selected neighborhood scale.

[0073] The specific processing method is: for a given grid point (x0, y0), the value of this grid point is determined by the values in the surrounding "neighborhood", and this definition method is extended to the entire space (x, y). The "neighborhood" constructed for each grid point is determined by the influence radius (or square area). The score of each grid point is the number of grid points in the "neighborhood" that exceed the defined threshold (for example, 0.1, 5, 10, 25 mm / 6 hours) divided by the total number of grid points in the "neighborhood". This score can be interpreted as the probability that the precipitation in the neighborhood equals or exceeds the threshold.

[0074] Embodiment 5:

[0075] This example is further designed based on the first embodiment in that the corrected typhoon path forecast and the mesoscale probability forecast after probability conversion are fused using a frequency matching method.

[0076] Example 6:

[0077] This example is further designed on the basis of Example 5 in that the frequency matching fusion method in this example includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast as the typhoon precipitation area forecast, and using the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion, and finally obtaining a precipitation forecast with better quality (the predicted precipitation area and magnitude are closer to the actual situation).

[0078] Embodiment seven:

[0079] This example is further designed on the basis of Example 5 in that the frequency matching fusion method in this example includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast and the mesoscale deterministic forecast as the typhoon precipitation area forecast, and using the mesoscale probabilistic forecast after probabilistic conversion as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion, and finally obtaining a precipitation forecast with better quality (the predicted precipitation area and magnitude are closer to the actual situation).

[0080] Embodiment 8:

[0081] The present invention provides a dual-resolution model typhoon forecast post-processing system, wherein the dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched, wherein the global-scale ensemble forecast includes a plurality of members, such as 51; the post-processing system includes a correction module, a probability conversion module, and a fusion module;

[0082] The correction module is used to correct the typhoon track forecast in the global-scale ensemble forecast in relation to the forecast quality by combining the historical typhoon data of the target area. That is, the typhoon track correction method is used to screen out members of the global-scale ensemble forecast with better quality and closer to the actual situation. Then, based on these members (the corresponding forecasts), a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed to improve the accuracy of the ensemble forecast.

[0083] The probability conversion module is used to convert the mesoscale deterministic forecast into a mesoscale probabilistic forecast;

[0084] The fusion module is used to fuse the revised typhoon path forecast with the mesoscale probability forecast.

[0085] Embodiment 9:

[0086] An electronic device of the present invention includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method described in any one of the embodiments.

[0087] A computer-readable storage medium of the present invention stores a computer program, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0088] Application examples:

[0089] This example simulates the post-processing method of the present invention for Typhoon Lekima No. 1909. Typhoon Lekima No. 1909 was generated on the sea east of the Philippines on August 4, 2019 (Beijing time, the same below). The Central Meteorological Observatory upgraded it from a tropical depression to a tropical storm at 17:00. Then, "Lekima" moved northwest under the guidance of the outer airflow of the subtropical high pressure in the western Pacific. At 05:00 on August 7, its intensity reached the typhoon level, and it was strengthened into a super typhoon at 23:00 on the same day. Since then, "Lekima" has maintained this intensity until it landed in the coastal area of Chengnan Town, Wenling City, Zhejiang Province in the early morning of August 10. At the time of landing, the maximum wind force near the center of the typhoon reached level 16 (52 meters per second), and the minimum pressure in the center was 930hPa. After landing, the trajectory of "Lekima" gradually turned from northwest to north. On the evening of the 10th, the typhoon weakened to a tropical storm and entered Jiangsu Province. After entering the Yellow Sea on the 11th, Lekima continued northward and made landfall again in Qingdao, Shandong Province. During this period, Typhoon Lekima maintained tropical storm strength. In the early morning of the 12th, Lekima crossed the Shandong Peninsula and entered Laizhou Bay, where it remained for a while before gradually weakening and dissipating.

[0090] Figure 2(a) shows the path of Typhoon Lekima from August 4 to 13, 2019. From August 9 to 12, Typhoon Lekima brought significant rainfall to eastern China, with the heaviest rainfall concentrated in Zhejiang, Jiangsu, and Shandong provinces. As shown in Figure 2(b), during the three days with the highest rainfall concentration, cumulative rainfall exceeded 100 mm at over 60% of rain gauges in East China. Lekima was the third strongest typhoon to make landfall in Zhejiang since 1949. After making landfall, it maintained tropical storm intensity or above for 44 hours, the sixth-longest duration since 1949. Its combined wind and rain intensity index was the highest since 1961.

[0091] Figure 3(a) shows the spatial distribution of accumulated precipitation from the global-scale ensemble forecast (EC-eps) from 08:00 on August 9 to 08:00 on August 12; Figure 3(b) shows the spatial distribution of accumulated precipitation from the mesoscale deterministic forecast (WARMS) from 08:00 on August 9 to 08:00 on August 12. Figures 3(a) and 3(b) show that the WARMS precipitation forecast provides a closer representation of the actual precipitation distribution and magnitude than the EC-eps. Due to its low horizontal resolution, the EC-eps forecast shows overly smoothed precipitation center boundaries and underestimates the magnitude of heavy rainfall. However, the high-resolution WARMS forecast has two significant problems: on the one hand, it overpredicts precipitation in southeastern China; on the other hand, after 20:00 on August 10, when Typhoon Lekima moved northward toward Shandong Province, the forecast underestimated the intensity and extent of the heavy rain in northeastern China. Overall, the low-resolution global model underestimated the precipitation magnitude while the location of the precipitation center was relatively accurate. The high-resolution WARMS forecast predicted a large amount of precipitation but the precipitation center deviated greatly from the actual situation.

[0092] This example verifies whether the correction of typhoon path forecast by combining subjective and objective member selection method is correct. In this example, the global scale ensemble forecast adopts ECMWF global model forecast. To compare the path error before and after correction, as shown in Figure 2, Figure 4 As shown in the figure, within the forecast periods of 24, 48, and 72, the average path error of the ensemble forecast after correction is smaller than the error before correction, indicating that the correction step has a positive effect on the global model path forecast of Typhoon Lekima.

[0093] This example further explores whether the newly formed ensemble forecast also has a positive correction on the precipitation forecast. Figure 5(a) to Figure 5(d)The RPS scores of the ensemble forecasts for precipitation before and after correction for different precipitation thresholds within a 72-hour forecast horizon (6-hour intervals) are shown. As the forecast horizon increases, the RPS scores of the models for precipitation predictability decrease at each precipitation threshold. Overall, the RPS values of the selected ensemble precipitation forecasts for each threshold are lower than those of the original ensemble forecasts, indicating that this member selection method also produces positive corrections to the precipitation forecast.

[0094] Figure 6 The figure shows the difference in RPS values between the original and revised ensemble precipitation forecasts for 72-hour forecast horizons (24-hour intervals) at different precipitation thresholds. As shown, the RPS difference is generally positive across most of East China. This means that the original ensemble forecast RPS is greater than the revised RPS across most regions, indicating a positive correction in the precipitation forecast. At each precipitation threshold, the area with negative RPS differences increases with increasing forecast horizon. In particular, at the heavy rain (10 mm / 6 hours) and torrential rain (25 mm / 6 hours) thresholds, the range of negative RPS differences in East China is even greater. Compared to light and moderate rain, the subjective and objective fusion member selection method performs less effectively for heavy and torrential rain at long forecast horizons.

[0095] This example also verifies the results of the post-processing method of the present invention. This example uses the precipitation forecast performance of the mesoscale WARMS forecast six hours before the onset time to determine the scale edge length of the neighborhood method based on the principle of maximizing the ROC area. Table 1 shows the scale edge lengths at different onset times under different precipitation thresholds, thereby obtaining the mesoscale probabilistic forecast WARMS-eps.

[0096] Table 1

[0097]

[0098] Then, the probability matching method is used to obtain the dual-resolution fusion precipitation forecast through the following two methods (Table 2).

[0099] Method 1 (hereinafter referred to as S1) uses the typhoon precipitation forecast (EC-eps average) in the revised global scale ensemble forecast as the typhoon precipitation area forecast, and uses the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then uses the frequency matching method to fuse them.

[0100] Method 2 (hereinafter referred to as S2) uses the typhoon precipitation forecast (EC-eps average value) in the revised global-scale ensemble forecast and the mesoscale deterministic forecast (WARMS) as the typhoon precipitation area forecast, and uses the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then uses the frequency matching method to fuse them.

[0101] Table 2

[0102] Project Name Forecast Type Precipitation area forecast Precipitation level forecast S1 Deterministic forecast EC-eps average WARMS-eps S2 Deterministic forecast EC-eps average + WARMS WARMS-eps

[0103] like Figure 7(a) to Figure 7(d) As shown in the figure, the ETS scores of S1 and S2 after the two dual-resolution fusion schemes are compared with the ECMWF ensemble forecast (average) and WARMS precipitation forecast before fusion. As can be seen from the figure, at the light rain, moderate rain, heavy rain and rainstorm thresholds, the ETS scores of S1 and S2 are greater than the ECMWF ensemble average and WARMS forecast, that is, the precipitation forecast after dual-resolution fusion correction is better than the low-resolution global model forecast and high-resolution mesoscale forecast before fusion. Figure 7a , b and Figure 7c As shown in Figure 2, the correction effect of the S1 and S2 precipitation forecasts on the ECMWF ensemble mean is more significant than that on the WARMS forecast at the light and moderate rain thresholds. In comparisons under different thresholds and forecast times, the overall forecast effect of S2 is slightly better than that of S1.

Claims

1. A dual-resolution model typhoon forecast post-processing method, characterized by: The dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched; the post-processing method includes: Combining historical typhoon data in the target area, the typhoon track forecast in the global-scale ensemble forecast is corrected in terms of forecast quality; Perform probability conversion on mesoscale deterministic forecasts to form mesoscale probabilistic forecasts; Fusing the revised typhoon track forecast with the mesoscale probability forecast; The specific method for correcting the forecast quality of the global-scale ensemble forecast by combining historical typhoon data of the target area includes: Based on the historical typhoon data of the target area, the members of the global-scale ensemble forecast are screened for their forecast quality. Based on the screened members, a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed; The method of screening the members of the global-scale ensemble forecast based on the historical typhoon data of the target area in terms of forecast quality and forming a revised typhoon track forecast and corresponding typhoon precipitation forecast based on the screened members is as follows: 1) Obtain the real-time location and subjective forecast of the typhoon impact process from historical typhoon data in the target area; 2) Based on the global scale ensemble forecast data, calculate the distance error between the path forecast of each member in the global scale ensemble forecast and the real-time position and subjective forecast during the typhoon impact process, and select the one with the smallest error. N members; 3) The arithmetic mean of the typhoon track forecasts of the N selected members is used as the revised typhoon track forecast and the corresponding typhoon precipitation forecast; The probability conversion of the mesoscale deterministic forecast is performed using the following method: The ROC curve areas of the mesoscale deterministic forecast at different scales are calculated, and the scale corresponding to the ROC curve with the largest area is selected as the neighborhood scale. The mesoscale deterministic forecast is then converted into a probability according to the selected neighborhood scale.

2. The dual-resolution model typhoon forecast post-processing method according to claim 1, characterized in that: The revised typhoon path forecast and the mesoscale probability forecast after probability conversion are fused using a frequency matching method.

3. The dual-resolution model typhoon forecast post-processing method according to claim 2, characterized in that: The frequency matching fusion method includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast as the typhoon precipitation area forecast, using the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion.

4. The dual-resolution model typhoon forecast post-processing method according to claim 2, characterized in that: The frequency matching fusion method includes using the typhoon precipitation forecast in the revised global-scale ensemble forecast and the mesoscale deterministic forecast as the typhoon precipitation area forecast, using the probabilistically converted mesoscale probability forecast as the typhoon precipitation magnitude forecast, and then using the frequency matching method for fusion.

5. A dual-resolution model typhoon forecast post-processing system, characterized by: The dual-resolution model typhoon forecast includes a global-scale ensemble forecast and a mesoscale deterministic forecast that are time-space matched; the post-processing system includes a correction module, a probability conversion module, and a fusion module; The correction module is used to correct the typhoon path forecast in the global-scale ensemble forecast in relation to the forecast quality in combination with the historical typhoon data of the target area; The probability conversion module is used to perform probability conversion on the mesoscale deterministic forecast to form a mesoscale probabilistic forecast; The fusion module is used to fuse the revised typhoon path forecast and the mesoscale probability forecast; The specific method for correcting the forecast quality of the global-scale ensemble forecast by combining historical typhoon data of the target area includes: Based on the historical typhoon data of the target area, the members of the global-scale ensemble forecast are screened for their forecast quality. Based on the screened members, a revised typhoon track forecast and corresponding typhoon precipitation forecast are formed; The method of screening the members of the global-scale ensemble forecast based on the historical typhoon data of the target area in terms of forecast quality and forming a revised typhoon track forecast and corresponding typhoon precipitation forecast based on the screened members is as follows: 1) Obtain the real-time location and subjective forecast of the typhoon impact process from historical typhoon data in the target area; 2) Based on the global scale ensemble forecast data, calculate the distance error between the path forecast of each member in the global scale ensemble forecast and the real-time position and subjective forecast during the typhoon impact process, and select the one with the smallest error. N members; 3) The arithmetic mean of the typhoon track forecasts of the N selected members is used as the revised typhoon track forecast and the corresponding typhoon precipitation forecast; The probability conversion of the mesoscale deterministic forecast is performed using the following method: The ROC curve areas of the mesoscale deterministic forecast at different scales are calculated, and the scale corresponding to the ROC curve with the largest area is selected as the neighborhood scale. The mesoscale deterministic forecast is then converted into a probability according to the selected neighborhood scale.

6. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images