1. A method of suppressing spurious convection, comprising the steps of:

determining the range of false convection:

determining the range in which the observed maximum radar echo is less than 10 dBZ and the difference between the maximum echo in the background field and the observed echo is greater than 10 dBZ as the range of the false convection;

secondly, determining the area where the radar echo is more than or equal to 0 and less than or equal to 10 dBZ as a non-precipitation echo area, and defining0w _max The average value of the maximum vertical speed of the background field non-precipitation echo area is obtained;

thirdly, assimilating three-dimensional radar echoes at first by using a method of integrated root-mean-square filtering, and then assimilating the echoes in a false convection area0w _max The value is circularly assimilated for four times to obtain an analysis field;

fourthly, continuously forecasting by utilizing the analysis field to obtain a forecasting field;

and fifthly, carrying out forecast inspection.

2. A method of suppressing spurious convection currents as defined in claim 1, wherein: step three, assimilating three-dimensional radar echo and assimilating two-dimensional radar echo in method for integrating root-mean-square filtering0w _max The time analysis variables include wind field, disturbance potential temperature, disturbance potential, water-vapor mixing ratio, rainwater mixing ratio, ice mixing ratio, aragonite mixing ratio, cloud mixing ratio, snow mixing ratio.

3. A method of suppressing spurious convection currents as defined in claim 2, wherein: the horizontal localization scheme and the vertical localization scheme adopted when assimilating three-dimensional radar echoes in the method for integrated root mean square filtering are GC functions.

4. A method of suppressing spurious convection currents as defined in claim 2, wherein: assimilating two-dimensional in the method of ensemble root-mean-square filtering0w _max The GC function is used as the horizontal localization function.

Background

The initial conditions in numerical weather forecast models (NWPs) play an important role in the skills of short-term weather forecasting. Radar observations, lightning data, satellite data, etc. have been assimilated by various methods to improve the initial field of NWPs, and these studies have yielded good results in terms of improved hit rates in strong convection processes. However, in real weather forecasting, a false forecast is often present. For fast moving convection systems, there may be significant positional deviation between the simulation and observation, when assimilating high spatial and temporal resolution observation data, two rain bands (one actual and one false) are likely to occur. Although the hit rate of precipitation is improved, the air report rate is still high. In still other cases, data assimilation may improve the prediction of strong convection processes, but also produce some false precipitation. Vendrasco et al (2016) show that assimilation of radar data using a three-dimensional mutation (3 DVAR) technique may produce spurious precipitation, as well as large errors in the location and amount of precipitation. These problems may be exacerbated in rapid cycle assimilation.

There have been some studies on how to suppress spurious convection. One reason for this may be the lack of a proper balance of dynamics and micro-physics in the initial analysis with the introduction of high spatio-temporal resolution data. Vendrasco et al (2016) attempted to minimize the prediction of overestimation by adding large-scale analytical constraints to the cost function, and Lin et al (2021) applied a large-scale constraint to the region model by spectral Nudging to suppress spurious precipitation. In addition to this, Reen (2017) believes that spurious convection may be due to noise, and a digital filtering method (Lynch, 1993) may be used to try to reduce it, however, digital filtering may also remove true atmospheric features. The methods can remarkably reduce the overestimated precipitation introduced in the data assimilation process, but have small inhibiting effect on false precipitation zone caused by inaccurate driving field.

In addition, some scholars have attempted to reduce the moisture or water content of spurious convection areas. Fierro et al (2019) assimilates the lightning density data to the moisture mass mix ratio for the lightning generating area and then removes the equivalent moisture mass throughout the field. The results show that these incremental negative adjustments of the moisture mass outside the lightning zone have negligible effect on spurious convection. The main reason is that only the water vapor at the initial moment is slightly adjusted, but the adjustment on the dynamic field and the thermal field is insufficient, new convection is still triggered in a false area, the water vapor in the surrounding area can be rapidly supplemented to an area with reduced water vapor under the action of the dynamic field, the water vapor in the initial field is removed by a physical initialization method, only the precipitation at the initial moment can be influenced, but the convection can be rapidly developed again at the subsequent moment.

Clear sky echo or rain-free echo can also be used for inhibiting false convection, and some obvious effects are achieved at present, but some problems still exist. First, the observation amount of the clear sky echo is large, and the assimilation workload is also large. Secondly, the detection accuracy of clear sky echoes is lower than precipitation echoes, and the observation error is not easy to be given (Rennie et al, 2018). Third, previous studies have shown that when the water-substance mixture ratio is used as a control variable for assimilation of reflectance data by 3d var, assimilation efficiency may be low due to small background reflectance and extremely large gradient of the cost function (Sun et al, 1997), while this problem is more pronounced when assimilating clear sky echoes, since the background reflectance in the clear sky echo region tends to be small (Kong, 2017).

In general, effective suppression of spurious convection requires consideration of both dynamic and thermal adjustments. On the one hand, it is desirable to thermally and dynamically suppress convection development in the pseudo-convection region. On the other hand, the heat and momentum exchange of the false convection zone with other surrounding areas needs to be attenuated, thereby limiting the regeneration of the false convection zone convection.

An ensemble square root filter (EnSRF) uses ensemble prediction to estimate flow-dependent background error covariance, which has been used to assimilate conventional observations, radar, satellite, and lightning data, etc. (Wang et al, 2015; Gao and Min, 2018). EnSRF has greater flexibility, and Gan et al (2021) propose a two-dimensional assimilation of lightning data conversion by EnSRFw _max The results show that two dimensions are assimilatedw _max Can effectively improve the thermal field, provide warm and humid environment for the development of convection, and adjust the dynamic field to ensure that the lower layer has strong convergence and the upper layer has strong divergence because of thew _max And the power and thermal variables of the modeHave strong correlation between them, are largerw _max Can be used to add information of medium and small scale to the background well and reduce the workload to a great extent.

Disclosure of Invention

The invention aims to provide a method for effectively reducing the forecast air report rate and inhibiting false convection.

In order to solve the above problems, the method for suppressing false convection according to the present invention comprises the following steps:

determining the range of false convection:

determining the range in which the observed maximum radar echo is less than 10 dBZ and the difference between the maximum echo in the background field and the observed echo is greater than 10 dBZ as the range of the false convection;

secondly, determining the area where the radar echo is more than or equal to 0 and less than or equal to 10 dBZ as a non-precipitation echo area, and defining0w _max The average value of the maximum vertical speed of the background field non-precipitation echo area is obtained;

thirdly, assimilating three-dimensional radar echoes at first by using a method of integrated root-mean-square filtering, and then assimilating the echoes in a false convection area0w _max The value is circularly assimilated for four times to obtain an analysis field;

fourthly, continuously forecasting by utilizing the analysis field to obtain a forecasting field;

and fifthly, carrying out forecast inspection.

Step three, assimilating three-dimensional radar echo and assimilating two-dimensional radar echo in method for integrating root-mean-square filtering0w _max Time analysis variables include wind field (u,v,w) Disturbance of the temperature (C:)prt) Disturbance potential (1)ph) Water-steam mixing ratio (qv) Mixing ratio of rainwater (1)qr) Mixing ratio of ice (C)qi) Mixing ratio of (a), (b), (c), (d)qg) Cloud mixing ratio (qc) Mixing ratio of snow (1)qs）。

The horizontal localization scheme and the vertical localization scheme adopted when assimilating three-dimensional radar echoes in the method for integrated root mean square filtering are GC functions.

Assimilating two-dimensional in the method of ensemble root-mean-square filtering0w _max The GC function is used as the horizontal localization function.

Compared with the prior art, the invention has the following advantages:

1. the invention is based on a WRF and EnSRF assimilation method of a mesoscale numerical mode, and assimilates in a false convection area0w _max The value is that the background field is adjusted from two aspects of a dynamic field and a thermal field, so that water vapor and ascending motion in the background field of the false convection region can be reduced, the development of convection of the region can be effectively inhibited, water vapor transmission from a peripheral region to the false convection region can be reduced, a ground cold pool is weakened, the regeneration and maintenance of false convection are not facilitated, the suppression effect on the forecast of false precipitation and false echo is obvious, and the forecast empty report rate can be effectively reduced.

2. Compared with the existing method for assimilating clear sky echoes or assimilating non-precipitation echoes, the method for assimilating clear sky echoes and the method for assimilating non-precipitation echoes provided by the invention in the false convection area0w _max The method has more obvious effect of inhibiting the false precipitation forecast.

3. Compared with the existing method for assimilating three-dimensional clear sky echoes or assimilating non-precipitation echoes, the method for assimilating two-dimensional echoes in the false convection area provided by the invention0w _max The method of values greatly reduces the amount of computation.

Drawings

The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.

FIG. 1 is a flow chart of the experimental design of the present invention, wherein RF, DA, w_maxRespectively representing radar echo, data assimilation and maximum vertical velocity, CTL, RDA and RDA _0Wmax respectively representing control test, test for assimilating radar echo, and simultaneous assimilation radar and0w _max testing of the values, Spin-up can be understood as: since the small and medium scale modes are started from the initial field without cloud, a certain time is needed for generating corresponding cloud water information, and the actual cloud water information is generated within a period of time after cold start, but the forecast within a few hours is inaccurate, and the hours are called as the "spin-up" time.

FIG. 2 is an assimilated radar echo and assimilated of the present invention0w _max And (3) a range. Wherein (a 1) is the maximum echo observed at 12 UTC on 27 th 6 th month in 2018, (a 2) is the maximum echo observed at 13 UTC on 27 th 6 th month in 2018, (a 3) is the maximum echo observed at 14 UTC on 27 th 6 th month in 2018, and (a 4) is the maximum echo observed at 15 UTC on 27 th 6 th month in 2018; (b1) is obtained by assimilating 12 UTC in 6 months and 27 days in 20180w _max (b 2) is the same range of 6/27/2018 UTC0w _max (b 3) is the same range of (1) in 2018, 6-month, 27-day, 14 UTC0w _max (b 4) is the same range of 6/27/15 UTC in 20180w _max The range of (1).

Fig. 3 is a diagram of the maximum radar echo and the predicted performance of the maximum echo predicted by the present invention. (a) The maximum radar echo of 16 UTC at 27.3.6.2018, (b) the maximum radar echo of 16 UTC at 27.27.2018 predicted by CTL test, (c) the maximum radar echo of 16 UTC at 27.27.27.2018 predicted by RDA test, (d) the maximum radar echo of 16 UTC at 27.27.3.2018 predicted by RDA _0Wmax test, (e) a predicted performance graph of 10 dBZ for the threshold, (f) a predicted performance graph of 20 dBZ for the threshold, and (g) a predicted performance graph of 30 dBZ for the threshold.

FIG. 4 is a graph of cumulative precipitation distribution and forecast performance of cumulative precipitation in accordance with the present invention. (a) The method comprises the following steps of (a) observing the cumulative precipitation of 15 UTC at 27 months and 6 months in 2018 to 18 UTC at 27 months and 6 months in 2018, (b) predicting the cumulative precipitation of 18 UTC at 27 months and 27 days in 6 months and 6 months in 2018 by a CTL test, (c) predicting the cumulative precipitation of 18 UTC at 27 months and 27 days in 6 months and 6 months in 2018 by a RDA test, (d) predicting the cumulative precipitation of 18 UTC at 27 months and 27 days in 6 months and 6 months in 6 months and 27 months in 2018 by a RDA _0Wmax test, (e) obtaining a prediction performance graph with a threshold value of 1 mm, (f) obtaining a prediction performance graph with a threshold value of 5 mm, and (g) obtaining a prediction performance graph with a threshold value of 10 mm.

Detailed Description

A method of suppressing spurious convection, comprising the steps of:

determining the range of false convection:

10 dBZ is considered as a threshold to distinguish precipitation echoes from non-precipitation echoes. The range where the observed maximum radar echo is less than 10 dBZ (including the observation default region) and the difference between the maximum echo in the background field and the observed echo is greater than 10 dBZ is determined as the range of spurious convection.

Secondly, determining the area where the radar echo is more than or equal to 0 and less than or equal to 10 dBZ as a non-precipitation echo area, and defining0w _max The average value of the maximum vertical speed of the background field non-precipitation echo area is obtained;

the non-precipitation echo zone is very weak to convection development, so the maximum vertical velocity in this region is a small value close to 0 m/s. In the invention, the background field is calculated according to 27-day example in 6-month-6-20180w _max Is 0.2 m/s.

A method of using integrated root mean square filtering (EnSRF) is adopted, and three-dimensional radar echoes are assimilated firstly>10 dBZ) and then assimilated in the spurious convection zone0w _max Values, the analytical field was obtained four times of cyclic assimilation.

The maximum vertical speed is a special two-dimensional variable, the EnSRF has strong flexibility, balance constraint is easy to establish between the maximum vertical speed and other variables, and in addition, the EnSRF method avoids sampling errors generated when disturbance observation is generated, so the method selects the EnSRF to assimilate the maximum vertical speed of lightning conversion.

Assimilation of three-dimensional radar echoes and assimilation of two-dimensional echoes in a method for integrated root-mean-square filtering0w _max Time analysis variables include wind field (u,v,w) Disturbance of the temperature (C:)prt) Disturbance potential (1)ph) Water-steam mixing ratio (qv) Mixing ratio of rainwater (1)qr) Mixing ratio of ice (C)qi) Mixing ratio of (a), (b), (c), (d)qg) Cloud mixing ratio (qc) Mixing ratio of snow (1)qs). Wherein: the horizontal localization scheme and the vertical localization scheme employed when assimilating three-dimensional radar echoes in the method of ensemble root mean square filtering are GC functions (Gaspari and Cohn, 1999). Assimilating two-dimensional in methods of ensemble root-mean-square filtering because vertical variations in vertical velocity in weak convection regions are insignificant0w _max The GC function is used as the horizontal localization function and the vertical localization function is not used.

And fourthly, continuously forecasting by utilizing the analysis field to obtain a forecasting field.

And fifthly, carrying out forecast inspection.

In order to verify the improvement effect of the method on the strong convection forecast, a strong convection process occurring in Jiangsu and Anhui areas of China for 27 days in 2018 is selected for carrying out example analysis, a group of control experiments and two groups of cyclic assimilation experiments are designed in total in the experiment, and a flow chart of the experiment design is shown in figure 1. RF, DA, w in FIG. 1_maxRespectively representing radar echo, data assimilation and maximum vertical velocity, CTL, RDA and RDA _0Wmax respectively representing control test, test for assimilating radar echo, and simultaneous assimilation radar and0w _max the test of (1). Spin-up can be understood as: since the small and medium scale modes are started from the initial field without cloud, a certain time is needed for generating corresponding cloud water information, and the actual cloud water information is generated within a period of time after cold start, but the forecast within a few hours is inaccurate, and the hours are called as the "spin-up" time.

The detailed experimental design is shown in table 1:

table 1 test design table

In the above-described experiment, experiment i (CTL) is a control experiment, experiment ii (RDA) is an experiment for assimilating a radar echo by the EnSRF method, assimilating times are 12, 13, 14, and 15 UTC in 6 and 27 months in 2018, and experiment iii (RDA — 0 Wmax) is an experiment for assimilating a radar echo and a false convection zone by the EnSRF method0w _max The test of (1). In the research, the WRF 4.0 version is selected, ensemble prediction of 21 global ensemble member prediction systems is used for providing an initial field and a boundary field for a WRF mode, in order to enhance sample dispersion, an RRTMG and Goddard radiation parameterization scheme is selected to disturb 21 samples into 42 samples, and in addition, the parameterized scheme is a Thompson micro-physical schemeMonin-Obukhov near-horizon solution, Mellor-Yamada-Janjic TKE planet boundary layer solution, unified Noah land solution, Tiedtke cloud parameterization solution (used only in D01). The study adopted two nested grids, the horizontal resolution of D01 was 9 km, the pattern grid was 171X 191, the horizontal resolution of D02 was 3 km, the pattern grid was 211X 271, and there were 50 layers in the vertical direction.

The invention assimilates radar echo and false convection area by adopting EnSRF scheme0w _max The values shown in FIG. 2 are the sum of the radar returns at four assimilation moments0w _max Distribution of values. Experiments (RDA) that assimilate only radar returns have relatively significant spurious convection, mainly located around the true convection.

Fig. 3 (a-d) and fig. 4 (a-d) show the distribution of radar maximum reflectivity and accumulated precipitation. 16 UTC, 27 th 6 th 2018, there are two strong echo regions: one at the junction of Anhui province and Jiangsu province and the other above the sea. The RDA test simulates well strong reflectivity, but the range of reflectivity is larger than the observed echo. The range of reflectivities predicted by the RDA — 0Wmax test is less than the RDA test. In the aspect of precipitation, the RDA experiment shows obvious false precipitation in the middle of Jiangsu, but the precipitation range simulated by the RDA _0Wmax experiment is close to the observed value.

Fig. 3 (e-g) and fig. 4 (e-g) are prediction performance graphs of radar echoes and accumulated precipitation with different thresholds, respectively, and the prediction performance graphs can reflect many pieces of information, including prediction deviation (FR, shown by a dotted line in the graph, closer to 1 indicates that the prediction deviation is smaller, greater than 1 indicates overestimation, and less than 1 indicates underestimation), hit rate (POD, higher indicates better prediction), critical success index (CSI, shown by a curve in the graph, a range of 0 to 1, approximately close to 1 is better), success rate (SR, 1 minus null report rate, closer to 1 is better), and overall, in the prediction performance graphs, approximately close to the upper right corner indicates more accurate prediction effect. The forecast performance graph of the reflectivity shows that the empty report rate of the RDA _0Wmax experiment is smaller than that of the RDA experiment, and the forecast performance graph of the precipitation can also obtain the same conclusion. In summary, assimilation in spurious convection regions by EnSRF0w _max Can effectively suppress falseAnd (4) convection.