A method and system for determining the water vapor source and transportation path of extreme precipitation

Abstract

The present invention discloses a method and system for determining the water vapor source and transportation path of extreme precipitation. The method includes obtaining extreme precipitation data and constructing an extreme precipitation model based on the extreme precipitation data; inputting the meteorological field simulated by the extreme precipitation model into a Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, and using the HYSPLIT model to determine M water vapor transportation trajectories of extreme precipitation; using a clustering analysis method to cluster the M water vapor transportation trajectories to obtain N water vapor transportation paths after clustering; calculating the water vapor contribution rates of the N water vapor transportation paths, and determining the main water vapor source of extreme precipitation according to the water vapor contribution rates. The present invention can not only quantify the distribution of water vapor in different stages and different regions, but also capture the vertical and horizontal mixing mechanisms of water from different sources above the atmospheric boundary layer, thereby ensuring the accuracy and precision of the determined water vapor source and water vapor transportation path.

Description

Technical Field

[0001] The present invention discloses a method and a system for determining the water vapor source and transportation path of extreme precipitation, belonging to the technical field of hydrological research. Background Art

[0002] Under the background of global warming, the frequency and intensity of extreme climate events in most land areas of the world are on the rise, resulting in frequent extreme hydrological events mainly including droughts and floods. Extreme precipitation is one of the most concerned and influential extreme climate events globally and is a research hotspot in the fields of extreme climate change and disaster prevention. The frequent occurrence of disasters such as floods and debris flows caused by extreme precipitation has had a serious impact on the ecological environment and the sustainable development of social economy.

[0003] The water vapor condition is one of the important factors for the formation of precipitation, and sufficient water vapor supply is a prerequisite for the formation of extreme precipitation. Analyzing the water vapor characteristics of extreme precipitation is of great significance for the research on the causes of extreme precipitation. The existing methods for determining water vapor transportation and its source cannot quantify the distribution of water vapor in different stages and different regions, and cannot capture the vertical and horizontal mixing mechanisms of water from different sources above the atmospheric boundary layer, resulting in inaccurate determination of the water vapor source and the water vapor transportation path, which affects the determination of extreme precipitation. Summary of the Invention

[0004] The purpose of this application is to provide a method and a system for determining the water vapor source and transportation path of extreme precipitation to solve the technical problem that the determined water vapor source and transportation path in the existing technology are inaccurate.

[0005] The first aspect of the present invention provides a method for determining the water vapor source and transportation path of extreme precipitation, including:

[0006] Obtain extreme precipitation data, and construct an extreme precipitation model according to the extreme precipitation data;

[0007] Input the meteorological field simulated by the extreme precipitation model into a hybrid single-particle Lagrangian integrated trajectory model, and use the hybrid single-particle Lagrangian integrated trajectory model to determine M water vapor transportation trajectories of the extreme precipitation;

[0008] Use the clustering analysis method to cluster the M water vapor transportation trajectories to obtain N clustered water vapor transportation paths;

[0009] Calculate the water vapor contribution rate of the N water vapor transportation paths, and determine the main water vapor source of the extreme precipitation according to the water vapor contribution rate.

[0010] Preferably, constructing an extreme precipitation model according to the extreme precipitation data specifically includes:

[0011] Construct the extreme precipitation model using a high - resolution regional climate model based on the extreme precipitation data.

[0012] Preferably, constructing the extreme precipitation model using a high - resolution regional climate model based on the extreme precipitation data specifically includes:

[0013] Determine the candidate parameters in the high - resolution regional climate model and construct multiple candidate extreme precipitation models;

[0014] Compare the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data to evaluate the simulation effect of the candidate extreme precipitation models;

[0015] Determine the extreme precipitation model according to the simulation effect.

[0016] Preferably, the candidate parameters include cumulus convection, microphysical processes, land surface processes, atmospheric long - wave radiation, short - wave radiation, and the planetary boundary layer.

[0017] Preferably, comparing the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data to evaluate the simulation effect of the candidate extreme precipitation models specifically includes:

[0018] Compare the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data;

[0019] Use the TS score, correlation coefficient, mean absolute error, and root mean square error to evaluate the simulation effect of the candidate extreme precipitation models.

[0020] Preferably, using the Hybrid Single - Particle Lagrangian Integrated Trajectory (HYSPLIT) model to determine M water vapor transport trajectories of the extreme precipitation, specifically including:

[0021] Determine the tracking parameters, where the tracking parameters include the initial height, initial position, start time, and tracking duration;

[0022] Input the tracking parameters into the HYSPLIT model to obtain M water vapor transport trajectories of the extreme precipitation.

[0023] Preferably, using the clustering analysis method to cluster the M water vapor transport trajectories to obtain N clustered water vapor transport paths, specifically including:

[0024] Obtain the average trajectory of the M water vapor transport trajectories;

[0025] Take each water vapor transport trajectory as a cluster and combine with the average trajectory to determine the spatial variance of each cluster;

[0026] Merge any two clusters into one cluster so that the total spatial variance of all clusters after merging is less than the total spatial variance before merging;

[0027] Obtain the change rate of the total spatial variance during the merging process, determine the clustering termination condition according to the change rate, and obtain N water vapor transport paths after clustering.

[0028] Preferably, determining the clustering termination condition according to the change rate specifically includes:

[0029] Judge whether the difference between the change rate of the current total spatial variance and the change rate of the previous adjacent total spatial variance is greater than a preset threshold. If so, the clustering terminates.

[0030] Preferably, calculating the water vapor contribution rate of N water vapor transport paths specifically includes:

[0031] Calculate the water vapor contribution rate of N water vapor transport paths according to the first formula, and the first formula is:

[0032]

[0033] In the formula, Qs is the water vapor contribution rate of a water vapor transport path, is the specific humidity value at the final position of the i-th water vapor transport trajectory included in this water vapor transport path, m is the total number of water vapor transport trajectories included in this water vapor transport path, is the specific humidity value at the final position of the j-th water vapor transport trajectory, and M is the total number of water vapor transport trajectories.

[0034] The second aspect of the present invention provides a system for determining the water vapor source and transport path of extreme precipitation, including:

[0035] A model construction unit, which is used to obtain extreme precipitation data and construct an extreme precipitation model according to the extreme precipitation data;

[0036] A trajectory determination unit, which is used to input the meteorological field simulated by the extreme precipitation model into the hybrid single-particle Lagrangian comprehensive trajectory model, and use the hybrid single-particle Lagrangian comprehensive trajectory model to determine M water vapor transport trajectories of the extreme precipitation;

[0037] A path determination unit, which is used to cluster the M water vapor transport trajectories by using a clustering analysis method to obtain N water vapor transport paths after clustering;

[0038] A source determination unit, which is used to calculate the water vapor contribution rate of N water vapor transport paths and determine the main water vapor source of extreme precipitation according to the water vapor contribution rate.

[0039] The method and system for determining the water vapor source and transportation path of extreme precipitation of the present invention have the following beneficial effects compared with the prior art:

[0040] The method and system for determining the water vapor source and transportation path of extreme precipitation of the present invention can not only quantify the distribution of water vapor in different stages and different regions, but also capture the vertical and horizontal mixing mechanisms of moisture from different sources above the atmospheric boundary layer, thereby ensuring the accuracy and precision of the determined water vapor source and water vapor transportation path, which is of great significance for understanding the formation and development of flood disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of the method for determining the water vapor source and transportation path of extreme precipitation provided by the present invention;

[0042] Figure 2 is a schematic structural diagram of the system for determining the water vapor source and transportation path of extreme precipitation provided by the present invention;

[0043] Figure 3 is the triple-nested region and superimposed terrain of the WRF model provided by the embodiment of the present invention;

[0044] Figure 4 is the spatial distribution map of the observed 24-hour cumulative precipitation during extreme precipitation in the embodiment of the present invention;

[0045] Figure 5 is the spatial distribution map of the simulated 24-hour cumulative precipitation during extreme precipitation in the embodiment of the present invention, where (a) to (l) are the spatial distribution maps of the cumulative precipitation corresponding to 12 parameterization schemes;

[0046] Figure 6 is the 72-hour backward trajectory map at different altitudes of UTC. Among them, (a) is the water vapor path result of the HYSPLIT model driven by WRF output data, and (b) is the water vapor path result of the HYSPLIT model driven by GDAS1 data;

[0047] Figure 7 is the evolution diagram of the specific humidity of the water vapor transportation trajectory over time. Among them, (a) is the water vapor path result of the HYSPLIT model driven by WRF output data, and (b) is the water vapor path result of the HYSPLIT model driven by GDAS1 data);

[0048] Figure 8 is the clustering result map of the 72-hour backward trajectory during extreme precipitation. Among them, (a), (b), and (c) are the results of the HYSPLIT model driven by the WRF model output data at 500m, 1500m, and 3000m respectively, and (d), (e), and (f) are the results of the HYSPLIT model driven by GDAS1 data at 500m, 1500m, and 3000m respectively.

[0049] In the figure, 101 is a model construction unit; 102 is a trajectory determination unit; 103 is a path determination unit; 104 is a source determination unit. Specific embodiments

[0050] In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present invention. However, those skilled in the art should clearly understand that the present invention can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from hindering the description of the present invention.

[0051] The method for determining the water vapor source and transportation path of extreme precipitation of the present invention has a process as Figure 1 shown, including:

[0052] Step 1: Obtain extreme precipitation data and construct an extreme precipitation model based on the extreme precipitation data. Specifically, it includes:

[0053] Step 1.1: Obtain extreme precipitation data;

[0054] Step 1.2: Construct an extreme precipitation model using a high-resolution regional climate model (Weather Research Forecast, abbreviated as WRF) based on the extreme precipitation data.

[0055] The above Step 1.2 specifically includes:

[0056] Step 1.2.1: Determine the candidate parameters in the high-resolution regional climate model (WRF) and construct multiple candidate extreme precipitation models.

[0057] The high-resolution regional climate model (WRF) used in the present invention is the latest generation of mesoscale numerical weather forecasting system. The candidate parameters in this model mainly include cumulus convection, microphysical processes, land surface processes, atmospheric longwave radiation and shortwave radiation, and the planetary boundary layer. For each parameter in WRF, multiple alternative schemes are provided. In addition, there are interactions between various physical processes. For example, the land surface transports surface fluxes (latent heat and sensible heat) to the planetary boundary layer, and the temperature, pressure, humidity, and wind in the planetary boundary layer affect the land surface process. The cumulus convection scheme affects the microphysical process through cloud roll-up effects and provides convective and non-convective precipitation to the land surface with the microphysical scheme respectively. The following are the physical processes mainly adopted in this article:

[0058] (1) Cumulus convection parameterization scheme

[0059] The main research objects of the cumulus convection parameterization scheme are the vertical fluxes (upward and downward airflows) and the compensating airflow motion outside the cloud, which can reasonably consider the latent heat release in the convective cloud and mainly calculate the sub-grid precipitation process. The cumulus convection parameterization schemes used in this application mainly include the Kain-Fritsch scheme and the Betts-Miller-Janjic scheme. The Kain-Fritsch scheme uses a simple cloud model accompanied by upward airflow, downward airflow, entrainment, and detrainment, including shallow convection processes and relatively rough microphysical processes. The Betts-Miller-Janjic scheme is adjusted and improved from the Betts-Miller scheme, with relaxation adjustment of the thermal profile, and the shallow convection process is the most important part of this scheme.

[0060] (2) Microphysical process parameterization scheme

[0061] The microphysical process parameterization scheme mainly describes the phase changes of water in the atmosphere. The Lin et al. scheme and the Kessler scheme are used in this application. The Kessler scheme comes from the COMMAS model. The Lin et al. scheme is a scheme with a relatively complex physical process description, including the prediction of six water substances: rainwater, cloud ice, cloud water, water vapor, snow, and graupel. This scheme is a relatively mature scheme in the WRF model and is more suitable for high-resolution simulations and theoretical research. The Kessler scheme is a simple warm cloud precipitation scheme that ignores the phase change process between liquid water and ice. The microphysical processes considered include the condensation of water vapor, the evaporation of rainwater, the collision of rainwater and cloud water, and the falling velocity of raindrops, etc.

[0062] (3) Land surface process parameterization scheme

[0063] The land surface process parameterization scheme provides the lower boundary conditions for the atmospheric dynamics equations to control and influence the climate development system. The Noah land surface process scheme is adopted in this application, which has four layers of soil moisture and humidity and includes the physical processes of snow cover and freezing.

[0064] (4) Atmospheric longwave radiation and shortwave radiation schemes

[0065] The longwave radiation scheme of this application adopts the RRTM scheme (Rapid Radiative Transfer Model), which uses a pre-determined look-up table to characterize the longwave radiation processes of gases such as water vapor, ozone, carbon dioxide, methane, etc., and the optical thickness of clouds, etc. The shortwave radiation scheme adopts the Dudhia scheme, which simply accumulates the solar radiation fluxes caused by clean air scattering, water vapor absorption, cloud reflection, and absorption.

[0066] (5) Planetary boundary layer scheme

[0067] The planetary boundary layer scheme characterizes the vertical sub-grid scale fluxes caused by eddy transport within the atmospheric column (not just the boundary layer). In this application, the YSU (Yonsei University) scheme and the MYJ (Mellor-Yamada-Janjic) scheme are mainly adopted. The YSU scheme is the second generation of the MRF (Medium Range Forecast Model) boundary layer scheme. Compared with the MRF scheme, the mixing layer height generated by thermal convection in the YSU scheme increases, while the mixing layer height generated by wind shear decreases. The MYJ uses the turbulence closure method of Mellor and Yamada to represent the turbulence above the surface layer. This scheme is applicable to the boundary layer under all stable conditions and weak unstable conditions, but has relatively large errors in the convective boundary layer.

[0068] Since the above individual candidate parameters include multiple parameterization schemes, this application adopts the method of combining multiple parameterization schemes to construct multiple candidate extreme precipitation models, and then determines a computational precipitation model with higher simulation accuracy from multiple candidate extreme precipitation models to ensure the accuracy of the subsequent determined water vapor sources and transport paths.

[0069] Step 1.2.2: Compare the simulated precipitation data output by the candidate extreme precipitation model with the extreme precipitation data to evaluate the simulation effect of the candidate extreme precipitation model, specifically including:

[0070] Compare the simulated precipitation data output by the candidate extreme precipitation model with the extreme precipitation data;

[0071] Use the TS score, correlation coefficient, mean absolute error, and root mean square error to evaluate the simulation effect of the candidate extreme precipitation model. Using the above multiple indicators to comprehensively evaluate the simulation effect of the candidate extreme precipitation model from both qualitative and quantitative perspectives can further ensure the accuracy of the simulation.

[0072] The TS score, correlation coefficient, mean absolute error, and root mean square error will be introduced in detail below.

[0073] (1) TS score

[0074] The TS score reflects the accuracy of effective precipitation forecasting in this application, and the score ranges from 0 to 1. When (TS = 1, the forecast completely coincides with the actual situation; when (TS = 0, the forecast completely does not match the actual situation.

[0075] The calculation formula of the TS score is as follows:

[0076]

[0077] Wherein, Na represents the number of stations with correct precipitation forecasts, that is, the number of times precipitation of this level appears in both observations and forecasts; Nb represents the number of false alarm stations, that is, precipitation of this level is forecast but not observed; Nc represents the number of missed alarm stations, that is, precipitation of this level is observed but not forecast.

[0078] The division of precipitation levels referred to in formula (1) is shown in Table 1.

[0079] Table 1 Division of precipitation levels for TS score

[0080]

[0081] (2) Correlation coefficient, mean absolute error and root mean square error

[0082] Since the cumulative precipitation distribution map can only qualitatively describe the precipitation distribution and precipitation intensity conditions, the present application uses error evaluation indicators to quantitatively compare and analyze the simulation effects of different parameterization schemes. The formulas for the mean absolute error MAE, root mean square error RMSE and correlation coefficient R of precipitation simulation values are as follows:

[0083]

[0084] Where, y iobserved is the actual observed value of precipitation at a certain grid point, is the mean value of the actual observed values of precipitation at multiple grid points, y ipredicted is the simulated value of precipitation at the grid point corresponding to this station, is the mean value of the simulated values of precipitation at multiple grid points, and i and N are a certain grid point and the total number of grid points in the region respectively.

[0085] The correlation coefficient R reflects the similarity between the simulated field and the observed field of the element.

[0086] The root mean square error RMSE reflects the deviation between the observed values and the simulated values in this region.

[0087] The mean absolute error MAE is the average of the absolute errors and can better reflect the actual situation of the simulation value error.

[0088] In the embodiments of the present invention, the candidate extreme precipitation models with higher TS scores and correlation coefficients, and lower root mean square errors and mean absolute errors are models with good simulation effects.

[0089] Specifically, thresholds for the TS score, correlation coefficient, root mean square error and mean absolute error can be preset in advance. When the TS score and correlation coefficient are greater than or equal to the corresponding thresholds, and at the same time the root mean square error and mean absolute error are less than the corresponding thresholds, the simulation effect of the candidate extreme precipitation model is better.

[0090] Step 1.2.3: Determine the extreme precipitation model according to the simulation results.

[0091] In the present invention, the candidate extreme precipitation model whose simulation results meet the preset conditions is used as the final extreme precipitation model. The preset conditions may be the best simulation results.

[0092] In the embodiment of the present invention, the extreme precipitation model includes not only the meteorological field corresponding to the extreme precipitation data, but also data such as terrain.

[0093] Step 2: Input the meteorological field simulated by the extreme precipitation model into the Hybrid Single Particle Lagrangian Integrated Trajectory Model, and use the Hybrid Single Particle Lagrangian Integrated Trajectory Model to determine M water vapor transport trajectories of the extreme precipitation.

[0094] In a specific embodiment, Step 2 is specifically: Input the data simulated and output by the extreme precipitation model with the best simulation results obtained in Step 1 into the Hybrid Single Particle Lagrangian Integrated Trajectory Model, and use the Hybrid Single Particle Lagrangian Integrated Trajectory Model to determine M water vapor transport trajectories of the extreme precipitation.

[0095] The Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT for short) in the embodiment of the present invention is a professional model for calculating and analyzing the transport and diffusion trajectories of air pollutants.

[0096] Since HYSPLIT can simulate the movement process of air particles at the selected time period and location, it can analyze and calculate the water vapor transport source areas of precipitation and the contribution rates of different water vapor channels. There are three projection methods for HYSPLIT, namely Mercator projection, Lambert projection, and polar projection. Its calculation requires meteorological data in a specific format (ARL format). A meteorological data file contains data for one or more time steps. Each time step data contains an ASCII index, which contains information such as layer, time, and variable. The three-dimensional variables of meteorological data required by the HYSPLIT model are shown in Table 2.

[0097] Table 2 Three-dimensional variables of meteorological data in the HYSPLIT model

[0098]

[0099] Assume that the trajectory of an air particle moves with the wind field. As it moves with the air current, its trajectory is the integral of the position vectors in time and space. The vector velocity at the position of the particle is obtained by linear interpolation in time and space. The specific tracking method is as follows:

[0100] P(t + Δt) = P(t) + 0.5[V(P, t) + V(P′, t + Δt)]Δt (5)

[0101] P′(t + Δt) = P(t) + V(P, t)Δt (6)

[0102] Wherein, P(t + Δt) is the final position of the air parcel, P′(t + Δt) is the initial imaginary position of the air parcel, Δt is the time step for tracking, V(P, t) is the three-dimensional velocity vector at the initial position, and V(P′, t + Δt) is the three-dimensional velocity vector of the initial imaginary position.

[0103] The HYSPLIT model calculates its advection and diffusion processes using the Lagrangian method. Before inputting the meteorological data into the HYSPLIT model, the meteorological data maintains its original format at the horizontal coordinates and is interpolated into the terrain-following vertical coordinate system during the calculation:

[0104]

[0105] Wherein, Z top is the top of the trajectory model coordinate system, Z mst is the lower boundary height of the coordinate, and Z gl is the terrain height.

[0106] The present invention inputs the meteorological field into the Hybrid Single-Particle Lagrangian Integrated Trajectory Model, even if the meteorological field in the model is the meteorological field obtained in step 1.

[0107] Wherein, determining M water vapor transport trajectories of extreme precipitation using the Hybrid Single-Particle Lagrangian Integrated Trajectory Model specifically includes:

[0108] Determining tracking parameters, where the tracking parameters include the initial height, initial position, start time, and tracking duration;

[0109] Inputting the tracking parameters into the Hybrid Single-Particle Lagrangian Integrated Trajectory Model to obtain M water vapor transport trajectories of extreme precipitation.

[0110] Step 3: Using the clustering analysis method to cluster the M water vapor transport trajectories to obtain N water vapor transport paths after clustering.

[0111] Since the number of obtained trajectories is large and it is impossible to accurately distinguish the distribution characteristics of each trajectory, in order to more intuitively and clearly understand the distribution of various trajectories, this article uses cluster analysis to cluster the trajectories. The basic idea of clustering is: clustering according to the principle of the closest spatial distance of the trajectories, and merging multiple trajectories with the most similar paths. Specifically includes:

[0112] Step 31: Obtaining the average trajectory of the M water vapor transport trajectories;

[0113] Step 32: Take each water vapor transport trajectory as a cluster, and in combination with the average trajectory, determine the spatial variance of each cluster. The airborne variance is the sum of the squares of the distances between each trajectory and the corresponding points on the average trajectory. The spatial variance of the trajectory at the starting point is 0, and each trajectory is a cluster at the initial moment.

[0114] Step 33: Combine any two clusters into one cluster so that the total spatial variance (TSV) of all clusters after combination is less than the total spatial variance before combination;

[0115] Step 34: Obtain the change rate of the total spatial variance during the combination process, determine the clustering termination condition according to the change rate, and obtain N water vapor transport paths after clustering.

[0116] Determining the clustering termination condition according to the change rate in this step specifically includes:

[0117] Judge whether the difference between the change rate of the current total spatial variance and the change rate of the previous adjacent total spatial variance is greater than a preset threshold. If so, the clustering terminates.

[0118] In the previous several trajectory combinations, the TSV first increases rapidly and then increases slowly. When clustering to a certain number of clusters, when continuing to cluster, the TSV increases rapidly again, indicating that the difference between the combined trajectories is large at this time, then the clustering terminates, and this point is taken as the end point of the clustering.

[0119] Step 4: Calculate the water vapor contribution rate of N water vapor transport paths, and determine the main water vapor sources of extreme precipitation according to the water vapor contribution rate.

[0120] Among them, calculating the water vapor contribution rate of N water vapor transport paths:

[0121]

[0122] In the formula, Qs is the water vapor contribution rate of a water vapor transport path, is the specific humidity value at the final position of the i-th water vapor transport trajectory included in this water vapor transport path, m is the total number of water vapor transport trajectories included in this water vapor transport path, is the specific humidity value at the final position of the j-th water vapor transport trajectory, and M is the total number of water vapor transport trajectories.

[0123] Then determine the main water vapor sources of extreme precipitation according to the water vapor contribution rate. It is mainly possible to select the water vapor source with the largest water vapor contribution rate or the water vapor source corresponding to a value greater than the preset water vapor contribution rate threshold as the water vapor source of extreme precipitation.

[0124] The second aspect of the present invention provides a system for determining the water vapor sources and transport paths of extreme precipitation, as Figure 2As shown in the figure, it includes: a model construction unit 101, a trajectory determination unit 102, a path determination unit 103, and a source determination unit 104.

[0125] Among them, the model construction unit 101 is used to obtain extreme precipitation data and construct an extreme precipitation model based on the extreme precipitation data;

[0126] The trajectory determination unit 102 is used to input the meteorological field simulated by the extreme precipitation model into the hybrid single-particle Lagrangian integrated trajectory model, and use the hybrid single-particle Lagrangian integrated trajectory model to determine M water vapor transport trajectories of extreme precipitation;

[0127] The path determination unit 103 is used to cluster the M water vapor transport trajectories by using the clustering analysis method to obtain N water vapor transport paths after clustering;

[0128] The source determination unit 104 is used to calculate the water vapor contribution rate of the N water vapor transport paths and determine the main water vapor sources of extreme precipitation according to the water vapor contribution rate.

[0129] The method and system for determining the water vapor source and transport path of extreme precipitation of the present invention can not only quantify the distribution of water vapor in different stages and different regions, but also capture the vertical and horizontal mixing mechanisms of water from different sources above the atmospheric boundary layer, so as to ensure the accuracy and precision of the determined water vapor source and water vapor transport path, which is of great significance for understanding the formation and development of flood disasters.

[0130] The present application will be described in detail below by taking the weather system in the Weihe River Basin as an example.

[0131] The WRF model uses 6-hour time intervals and 1°×1° NCEP-FNL data as its initial field and lateral boundary conditions, and applies three-layer two-way nesting. Among them, the coarse grids d01 and d02 simulate the large-scale environmental field, with horizontal resolutions of 27 Km and 9 Km respectively, and the number of grid points are 121×113 and 151×133 respectively. The fine grid d03 mainly covers the weather system in the Weihe River Basin, with a horizontal resolution of 3 Km and the number of grid points of 265×187, as Figure 3 shown. The model has 35 layers in the vertical direction, and the top pressure of the model is 50 hPa. The selected time for extreme precipitation in this embodiment is from 00:00 on July 23, 2010 to 00:00 on July 25, 2010. The model results are output every 3 hours in the outer layer and every 1 hour in the innermost layer. The first 12 hours of the model are used as the'spin-up' time, and the time period that this application focuses on is from 12:00 on July 23, 2010 to 12:00 on July 24, 2010 (UTC).

[0132] The physical parameterization schemes adopted in this application mainly include the Goddard long-wave radiation scheme, the Dudhia short-wave radiation scheme, the Noah land surface process scheme, and the Monin-Obukhov surface layer scheme. Different cumulus convection parameterization schemes, boundary layer parameterization schemes, and microphysical process schemes in the WRF model have a great impact on the simulation results of the model. Different combinations of parameterization schemes are suitable for different regions. Since the resolution of the innermost nesting area is 3 Km, the cumulus convection parameterization scheme is not selected for the innermost nesting domain. In this application, 12 possible combinations of parameterization schemes are selected from different parameterization schemes of the WRF model, and combined with the observed precipitation data to determine the parameterization scheme suitable for the Weihe River Basin.

[0133] Several research results show that among all the schemes, the microphysical process scheme and the cumulus convection parameterization scheme have the greatest impact on the precipitation simulation results. WRFV4.1 provides 13 microphysical process schemes and 9 cumulus process parameterization schemes. Combining the physical meanings and applicable conditions of different schemes, this application selects 2 microphysical process schemes, 3 cumulus convection process parameterization schemes, and 2 boundary layer parameterization schemes for the Weihe River Basin for scheme combination, in order to analyze the influence of different schemes of WRFV4.1 on the precipitation simulation accuracy and parameter sensitivity, and then optimize the best precipitation simulation parameterization scheme. The alternative schemes for the microphysical process include: the Kessler scheme, the Lin et al. scheme. The alternative schemes for the cumulus convection parameterization process include: the shallow convection Kain-Fritsch (new Eta) scheme, the Betts-Miller-Janjic scheme, and the Grell-Freitas scheme. 12 alternative parameterization scheme combinations are determined, as shown in Table 3.

[0134] Table 3 12 Parameterization Scheme Combinations for the Weihe River Basin

[0135]

[0136] The spatial distribution of the observed cumulative precipitation from 12:00 on July 23 to 12:00 on July 24, 2010 (UTC) is as Figure 4 shown. The precipitation mainly occurs in the central and southern parts of Shaanxi and the eastern part of Gansu. The precipitation shows a trend of gradually increasing from west to east, weakening and then strengthening. The high-value area is located in the southeastern part of Shaanxi, and the 24-hour cumulative precipitation reaches 221.4 mm. Given that the spatial resolution of WRF is 3 Km, its spatial distribution is more detailed than the grid observation data with a resolution of 0.1°. From the perspective of the simulation of the precipitation spatial distribution in the entire Weihe River Basin, all schemes can generally simulate the precipitation range in the Weihe River Basin during this time period. The simulation of the total precipitation range and distribution characteristics is generally the same as the observed situation, but the simulation of the rainstorm center and rainstorm magnitude is slightly insufficient.

[0137] By comparing the simulated precipitation results of different parameterization schemes with the measured precipitation, the advantages and disadvantages of the simulation results can be intuitively reflected. The precipitation simulation results of 12 parameterization schemes are as follows Figure 5 shown. Compared with the measured precipitation, among the 12 parameterization scheme combinations, Scheme 5 is the most consistent with the actual situation in terms of rainfall distribution, Scheme 7 is the most consistent in terms of rainfall magnitude, and Scheme 2 has a large deviation from the measured precipitation data in terms of rainfall magnitude and rainfall area. The same microphysical process and boundary layer parameterization scheme are adopted in Experimental Schemes 1, 3, and 5, and only the cumulus convection parameterization scheme is set differently. The precipitation distribution simulated by the Kain-Fritsch scheme adopted in Scheme 1 is relatively large, and there is a large deviation from the actual situation in the precipitation area. And there are multiple heavy rain centers, but the simulation of precipitation magnitude is good. Scheme 3 adopts the Betts-Miller-Janjic scheme, and the precipitation distribution simulated by this scheme is relatively concentrated, and the center is roughly located at 33.7°N, 109.1°E, which is close to the actual center position. Scheme 5 adopts the Grell-Freitas scheme, and the simulation of the precipitation area and precipitation center by this scheme is the closest to the actual situation, but the precipitation center value simulated by Scheme 5 is on the large side. When the same microphysical process and cumulus convection parameterization scheme are adopted and only the boundary layer scheme is set differently, it can be found that the precipitation area and magnitude of the precipitation simulation are roughly consistent, indicating that the boundary layer scheme has little influence on the precipitation simulation this time. Among the 12 schemes, the first two pairs of schemes have the same cumulus convection parameterization scheme and boundary layer parameterization scheme, and only the microphysical process scheme is set differently. From Figure 5 it can be qualitatively judged that compared with the Kessler scheme, the Lin scheme is more suitable for the short-term extreme precipitation simulation in the Weihe River Basin in terms of precipitation area and precipitation magnitude. Different boundary layer parameterization schemes do not show large differences in this simulation, so it is necessary to quantitatively evaluate the advantages and disadvantages of each combination scheme according to the evaluation index.

[0138] In this application, the TS score, correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE) are used to evaluate the precipitation simulation results for the 24-hour cumulative precipitation simulated by the WRF model. The TS score results are shown in Table 4. The correlation coefficient, root mean square error (RMSE), and mean absolute error (MAE) results are shown in Table 5.

[0139] It can be seen from Table 4 and Table 5 that as the precipitation level increases, the TS score decreases rapidly. Among them, the light rain level is about 0.60, and the TS scores of moderate rain and above levels are all below 0.41. It can be seen that each experimental scheme has the best simulation for light rain. Among the TS scores of each rainfall level, the precipitation simulation effect of the microphysical process scheme using the Lin scheme is slightly better than that of the Kessler scheme. The Lin scheme is a relatively complex microphysical scheme in the WRF model and is more suitable for high-resolution simulation. When the cumulus convection parameterization scheme uses Grell-Freitas, the TS score performs well. The parameter combination schemes 5 and 11 have the highest TS scores for light rain, but are slightly insufficient for the simulation of heavy rain and above levels. Schemes 3 and 9 have the highest TS scores for extremely heavy rain, indicating that the Lin-BMI combination scheme is more suitable for the simulation of high-level precipitation in this region. In addition, schemes 4, 10, and 12 have the worst simulation effect for extremely heavy rain, and the TS score is 0.

[0140] Table 4 TS scores of WRF simulated precipitation in the Weihe River Basin

[0141]

[0142] Table 5 Calculation results of evaluation indicators for simulated precipitation of different parameterization scheme combinations

[0143]

[0144] Similar to the TS score results, the correlation coefficient indicators of schemes 3, 5, 9, and 11 are relatively high, and the mean absolute error and root mean square error are the smallest, showing an absolute advantage. That is, when the microphysical process scheme uses the Lin scheme, the correlation between the simulated precipitation result and the measured precipitation result is higher. Scheme 3 has the highest TS score in the extremely heavy rain level, and its TS scores in the heavy rain and below levels are also good. Considering the results of R, MAE, and RMSE in Table 5 comprehensively, the scheme combination of scheme 3, that is, the microphysical process uses the Lin scheme, the cumulus convection parameter uses the Betts-Miller-Janjic scheme, and the boundary layer scheme uses the YSU scheme, has the best simulation effect on the extreme precipitation in the Weihe River Basin here.

[0145] The parameter schemes in the adopted extreme precipitation model are shown in Table 6.

[0146] Table 6 Parameterization schemes of the WRF model

[0147]

[0148] In most cases of HYSPLIT water vapor tracking, the meteorological input data are generated by the Global Data Assimilation System (GDAS), with a horizontal resolution of 10×1° or 0.5°×0.5°. The temporal resolution is 3 hours. GDAS data can qualitatively describe the general characteristics of weather conditions, but the coarse data resolution in horizontal, vertical, and temporal scales may sometimes introduce uncertainties in three-dimensional trajectory simulations. Given the lack of observational data on trajectories, it is impossible to verify the reliability of the results by comparing the simulated trajectory paths with the observational data, and existing studies have shown that there are certain differences in the results of water vapor transport characteristics calculated using different datasets. To improve the reliability of the results, in this study, the HYSPLIT model was driven by the WRF model run results and the GDAS data with a spatial resolution of 1°×1° and a temporal resolution of 6 hours for geopotential height, temperature, zonal wind, and meridional wind at 23 levels from 1000 to 20 hPa, respectively, to explore the feasibility of driving the HYSPLIT model with the WRF model run results and the water vapor sources and main paths during heavy precipitation processes in the Weihe River Basin. The physical parameterization scheme adopted by the WRF model is shown in Table 6. The d01 domain was selected as the backward trajectory simulation area, with a horizontal resolution of 27 km and a model top height of 50 hPa. The WRF output results were converted into the input format (arl format) of the HYSPLIT model, and finally, the one-way offline coupling of the WRF model and the HYSPLIT model was achieved.

[0149] Since more than 90% of the water vapor in the Chinese region is concentrated in the atmosphere below 500 hPa, three altitude layers corresponding to 3000 m, 1500 m, and 500 m at 700 hPa, 850 hPa, and 925 hPa were selected as the initial simulation altitudes in this paper. First, the three-dimensional motion trajectories were traced backward for 72 hours with the time and location of the maximum hourly cumulative precipitation as the initial time and initial point, and the trajectory point positions and hourly physical quantity fields (altitude, specific humidity) were output every 1 hour. Secondly, all trajectory points were traced backward for another 72 hours every 1 hour. Based on the HYSPLIT model, the position of the trajectory can be clearly seen for a single initial position, but in the case of multiple levels and a large number of trajectories, it is not easy to quantitatively describe the contributions of different moisture sources. Therefore, based on the clustering analysis of the method itself, the trajectories were classified by analyzing the change of TSV (total spatial variance). By analyzing the growth rate of the spatial variance, it was found that the growth rate of the variance during the clustering process of the trajectories increased rapidly after the clustering result was less than 2 or 3, and thus the final clustering number of the simulated trajectories was determined.

[0150] To understand the water vapor transport path during this rainstorm process, the HY SPLIT model based on the Lagrangian method was used to trace the water vapor source of the extreme precipitation process from July 22 to 24, 2010. The time of the maximum hourly cumulative precipitation, that is, 22:00 (UTC) on July 23, 2010, was selected as the starting time for tracing the water vapor source. The longitude and latitude coordinates of the initial points were 33.75°N, 110.5°E and 34.8°N, 107.7°E respectively, and backward tracking was carried out for 72 hours at three altitude levels.

[0151] It can be seen from Figure 6 that the trajectories obtained by driving the HYSPLIT model with data of different resolutions are not the same. During the extreme precipitation in this embodiment, among all the trajectories, only the WRF-HYSPLIT at the 3000m altitude has a northward water vapor trajectory, and its source areas are located within Mongolia. While the water vapor of GDAS1-HYSPLIT at the 3000m altitude comes from within Thailand, passes through the South China Sea of our country, enters China, and reaches the study area, which is a southward water vapor trajectory. The typhoon "Chanthu" landed along the coast of Guangdong on July 22, 2010, and the extreme precipitation in the Weihe River Basin was severely affected by this, so its water vapor mostly comes from the southward water vapor channel. In addition, from the height evolution of the air parcel trajectories at each layer, except for the 3000m water vapor trajectory from the northward channel, the rest of the water vapor trajectories come from the lower troposphere below 2000m altitude, indicating the height of the source water vapor for this extreme precipitation. The 3000m water vapor trajectory of the northward channel comes from the troposphere above about 6000m, and its specific humidity always remains at about 3g / kg, having little impact on this extreme precipitation. In addition, when the rainstorm occurs, except for the 3000m water vapor trajectory, there is an obvious uplifting process in both simulation schemes, which is caused by the vertical upward movement during the rainstorm. The transport height of the trajectories generated by WRF-HYSPLIT is slightly higher than that of GDAS1-HYSPLIT. Therefore, the specific humidity values of the spatial mass points on each trajectory of WRF-HYSPLIT are lower than those of GDAS1-HYSPLIT. As Figure 7 shown, the specific humidity decreases with the increase of height in the vertical height.

[0152] When driving the HYSPLIT model with meteorological fields of two different resolutions, due to the terrain uplifting effect, the simulated air mass height increases with the increase of the terrain height and decreases with the decrease of the terrain height. Although the water vapor paths come from different altitude levels, the trend of the water vapor trajectories from the same direction is generally the same. However, since the resolution of the WRF model output data is much higher than that of the GDAS1 data, it has a stronger ability to capture information on the meso-scale and can more accurately describe the terrain height where the air mass is located to a certain extent, thus resulting in Figure 6There is a northerly water vapor path that exists in WRF-HY SPLIT but not in GDAS1-HYSPLIT, indicating that using driving data with different resolutions has a significant impact on the trajectory calculation results.

[0153] To further quantify the water vapor transport conditions at different heights during the extreme precipitation process, all trajectory points are re-traced backward every 1h for 3d to obtain the trajectory distribution of the same initial points at different times during this extreme precipitation process, and then the water vapor migration path during the entire precipitation process is analyzed.

[0154] Different water vapor transport channels correspond to different water vapor transport heights. The trajectories obtained by cluster analysis at 3 heights in this extreme precipitation case example are as Figure 8 shown. For the 500m water vapor transport trajectory in WRF-HYSPLIT (hereinafter referred to as W-H), a total of 3 paths appear ( Figure 8 (a) in). Path 1 is a southerly water vapor transport, transporting northward from a height of 1500m. The air mass height gradually drops to 500m at 16:00 on the 21st, and the number of trajectories accounts for 17%. Paths 2 and 3 both come from the southeast direction, accounting for 21% and 52% of the total number of trajectories respectively. The initial transport heights are 500m and 1000m respectively, and the trajectory heights fluctuate greatly. The 500m water vapor transport path of the HYSPLIT model driven by GDAS1 data (hereinafter referred to as G-H) is as Figure 8 (d) in. The three water vapor transport paths all come from the southeast direction, but the starting points are different. The three water vapor transport channels account for 59%, 22%, and 19% of the total number of trajectories respectively. The first water vapor transport path is transported from near the ground, and the air flow migration height is relatively low. The other two water vapor transport paths both start transporting from about 500m, and the migration height remains almost unchanged.

[0155] Comparatively analyzing the average trajectories obtained by cluster analysis of the W-H and G-H extreme precipitation cases at a height of 1500m ( Figure 8 (b) and (e) in), the water vapor channels at this height are more diverse than those at a height of 500m. There are three water vapor paths in W-H. Paths 1 and 2 are similar to the W-H 500m water vapor paths, both coming from the southwest direction, accounting for 43% and 47% of the total number of trajectories respectively, and the initial transport heights are the same, both 1500m. Path 3 accounts for a relatively small proportion, only 10%. It starts transporting from the northeastern region of China, passes through Hebei and Shanxi provinces and enters the study area, and the water vapor transport height is much higher than that of Paths 1 and 2, starting from a height of 3000m and gradually decreasing to a height of 1500m over time. The G-H water vapor transport path is as Figure 8As shown in (e), there are three main water vapor transport channels, all originating from the South China Sea, carrying a large amount of warm and humid ocean water vapor, and passing through Guangdong, Hunan and other places in the northwest direction into the study area. Both Path 1 and Path 3 transport northward from near the 500m height in the South China Sea, but the channel 3 reaches the high value near 2000m after 48h of transportation and finally smoothly transitions to 1500m. Path 2 originates from the sea area near the 1200m height of the Nansha Islands in China, with a small water vapor contribution.

[0156] Comparative analysis of the average trajectories obtained from the clustering analysis of extreme precipitation cases of W-H and G-H at 3000m height ( Figure 8 in (c) and (f)) There are three main water vapor paths for W-H at 3000m. Path 1 enters the rainstorm area from Mongolia through Inner Mongolia, Shanxi and other provinces in China. This channel accounts for 25% of the total number of trajectories. Path 2 and Path 3 both come from the near southwest direction of the rainstorm area, both accounting for 38%. All three water vapor transport paths start transporting from a height of about 2500m, but the water vapor path from the Mongolian direction is much higher than the other two water vapor channels from the southwest direction, and its high value is near 4000m. The G-H water vapor transport path is as Figure 8 shown in (f). There are two water vapor channels both coming from the southwest direction. Path 1 starts transporting from the 1500m height in Thailand, passes through Laos, Vietnam and other places and enters the Chinese territory and finally reaches the rainstorm area, accounting for 53% of the total number of trajectories. Path 2 originates from near the 1000m height in Vietnam, passes through Hainan, Guangdong and other places and enters the study area, with a proportion of 47%. These two water vapor paths both come from near the 1200m height. As the air flow moves, the moving height gradually increases until it reaches the 3000m height.

[0157] In general, the water vapor for this extreme precipitation mainly comes from the southerly passage in the lower troposphere and the northerly passage in the middle and lower troposphere. Among them, the southerly water vapor passage accounts for the vast majority of the water vapor transport for this extreme precipitation. The water vapor in this southerly passage is closely related to the typhoon "Chanthu". "Chanthu" is located in the South China Sea of China and made landfall along the coast of Guangdong Province on the 22nd. Under the action of its peripheral circulation, the water vapor is continuously transported northward to the rainstorm area. The W-H and G-H water vapor paths at a height of 500 m in the lower troposphere are both the same, showing a southerly water vapor path. Compared with G-H, which only has a southerly water vapor path at heights of 1500 m and 3000 m, there are smaller proportions of northerly water vapor paths (10% and 25% respectively) in the W-H at heights of 1500 m and 3000 m. The reason for this difference may be that compared with the GDAS1 data, the higher-resolution output data of WRF driving HYSPLIT can generate more refined resolution trajectories to capture weak weather conditions. When the spatial scale of the weather process is less than 110 km and greater than 27 km, the output data of the WRF model can capture the wind field changes of this weather process, thus more accurately simulating the movement trajectory of the water vapor mass. However, the important details in these complex weather conditions are often ignored in the trajectories generated by the GDAS1 data. Different driving data have different description capabilities for the terrain height where the air particles are located. Therefore, the height difference in the transmission of the water vapor mass between the two may be related to the fact that the WRF model output data has a higher resolution and can more accurately identify complex terrains. In addition, the WRF output data has more vertical layers than the GDAS data. Therefore, the WRF output data can provide more height information for the HYSPLIT model in the vertical height.

[0158] The above are only several embodiments of the present application and do not impose any form of limitation on the present application. Although the present application is disclosed above with preferred embodiments, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications by using the disclosed technical content within the scope of the technical solution of the present application, which are equivalent to equivalent implementation cases and all belong to the scope of the technical solution.

Claims

1. A method for determining the water vapor source and transportation path of extreme precipitation, characterized in that, Including: Obtain extreme precipitation data and construct an extreme precipitation model based on the extreme precipitation data; Input the meteorological field simulated by the extreme precipitation model into a Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, and use the HYSPLIT model to determine M water vapor transport trajectories of the extreme precipitation; Cluster the M water vapor transport trajectories using a clustering analysis method to obtain N clustered water vapor transport paths; Calculate the water vapor contribution rates of the N water vapor transport paths, and determine the main water vapor sources of the extreme precipitation based on the water vapor contribution rates; Construct an extreme precipitation model based on the extreme precipitation data, specifically including: Construct the extreme precipitation model using a high-resolution regional climate model based on the extreme precipitation data, specifically including: Determine the candidate parameters in the high-resolution regional climate model and construct multiple candidate extreme precipitation models; Compare the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data to evaluate the simulation effects of the candidate extreme precipitation models; Determine the extreme precipitation model based on the simulation effects; Calculate the water vapor contribution rates of the N water vapor transport paths, specifically including: Calculate the water vapor contribution rates of the N water vapor transport paths according to the first formula, and the first formula is: ， In the formula, is the water vapor contribution rate of a water vapor transport path, is the specific humidity value at the final position of the th water vapor transport trajectory included in this water vapor transport path, is the total number of water vapor transport trajectories included in this water vapor transport path, is the specific humidity value at the final position of the th water vapor transport trajectory, is the total number of water vapor transport trajectories.

2. The method according to claim 1, characterized in that, The candidate parameters include cumulus convection, microphysics, land surface processes, atmospheric longwave radiation, shortwave radiation, and the planetary boundary layer.

3. The method according to claim 1, characterized in that, Compare the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data to evaluate the simulation effects of the candidate extreme precipitation models, specifically including: Compare the simulated precipitation data output by the candidate extreme precipitation models with the extreme precipitation data; Use the Threat Score (TS), correlation coefficient, mean absolute error, and root mean square error to evaluate the simulation effects of the candidate extreme precipitation models.

4. The method according to claim 1, wherein Use the HYSPLIT model to determine the M water vapor transport trajectories of the extreme precipitation, specifically including: Determine the tracking parameters, and the tracking parameters include the initial height, initial position, start time, and tracking duration; Input the tracking parameters into the HYSPLIT model to obtain the M water vapor transport trajectories of the extreme precipitation.

5. The method according to any one of claims 1-4, characterized in that, Cluster the M water vapor transport trajectories using a clustering analysis method to obtain N clustered water vapor transport paths, specifically including: Obtain the average trajectory of the M water vapor transport trajectories; Take each water vapor transport trajectory as a cluster, and combine it with the average trajectory to determine the spatial variance of each cluster; Merge any two clusters into one cluster so that the total spatial variance of all clusters after merging is less than the total spatial variance before merging; Obtain the change rate of the total spatial variance during the merging process, determine the clustering termination condition based on the change rate, and obtain the N clustered water vapor transport paths.

6. The method according to claim 5, characterized in that, Determine the clustering termination condition based on the change rate, specifically including: Judge whether the difference between the change rate of the current total spatial variance and the change rate of the previous adjacent total spatial variance is greater than a preset threshold. If so, the clustering terminates.

7. A system for determining the water vapor source and transportation path of extreme precipitation based on the method according to any one of claims 1-6, characterized in that, Including: A model construction unit, which is used to obtain extreme precipitation data and construct an extreme precipitation model according to the extreme precipitation data; A trajectory determination unit, which is used to input the meteorological field simulated by the extreme precipitation model into a hybrid single-particle Lagrangian comprehensive trajectory model, and use the hybrid single-particle Lagrangian comprehensive trajectory model to determine M water vapor transport trajectories of the extreme precipitation; A path determination unit, which is used to cluster the M water vapor transport trajectories by using a clustering analysis method to obtain N water vapor transport paths after clustering; A source determination unit, which is used to calculate the water vapor contribution rates of the N water vapor transport paths and determine the main water vapor sources of the extreme precipitation according to the water vapor contribution rates.

Images