New method for constructing regional weighted average temperature model considering multi-factor product

Abstract

The invention relates to the technical field of meteorology atmospheric precipitable water amount inversion, and discloses a novel regional weighted average temperature model construction method considering multiple factor products, which comprises the following steps of: utilizing radio sounding station observation data in a Yunnan and Guizhou plateau region, firstly preprocessing the data, eliminating obvious abnormal values and processing missing values, and then obtaining a new regional weighted average temperature model; the air pressure type data is vertically layered and standardized, so that the data formats of different sounding stations are consistent; the method comprises the following steps of: calculating an accurate atmospheric weighted average temperature through a numerical integration method by utilizing observation data and surface data of a sounding station in a region, and providing correlation factors for subsequent modeling according to the strength of correlation for screening key factors of modeling and analyzing linear correlation between the atmospheric weighted average temperature and surface temperature, latitude, elevation and the like; and based on a correlation analysis result, constructing a nonlinear regression model considering a multi-factor product by utilizing sounding station data of the region.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological atmospheric precipitable water inversion, and in particular to a new method for constructing a regional weighted average temperature model taking into account multi-factor products. Background Art

[0002] Water vapor plays a crucial role in the global water cycle, climate change, and the evolution of severe weather events. Therefore, analyzing the temporal and spatial trends of atmospheric water vapor is of great significance. As a novel means of detecting water vapor, the global navigation system (GNSS) is increasingly being used for water vapor detection due to its unique advantages, including all-weather observations, global coverage, high precision, and high temporal and spatial resolution. When deriving PWV from GNSS data, the water vapor conversion parameter π is used to convert ZWD into atmospheric water vapor. This parameter, primarily influenced by the weighted mean temperature of the atmosphere, directly determines the accuracy and reliability of PWV inversion using GNSS.

[0003] T m The data can be obtained through mathematical calculation based on sounding data or layered meteorological data, but it is affected by factors such as time and space. This method is not convenient for people to obtain more accurate T m At present, there are two main types of atmospheric weighted average temperature calculation models: those that require measured meteorological parameters and those that do not. The most representative of the former is the Bevis model, whose basic principle is to establish the atmospheric weighted average temperature and the surface temperature T s The linear model (T m =70.2+0.72T s ), this model is simple and practical, but it is difficult to obtain high-precision T under complex terrain and climate characteristics. m The most representative of the latter is the GPT3 model, the T of non-meteorological parameters m Although the model does not depend on the measured meteorological parameters, it is consistent with the T m There are certain systematic errors in the model, and its accuracy in local areas needs further verification.

[0004] In summary, the current T m The empirical model uses only a single T m The influencing factors only consider the characteristics of temporal and spatial factors, which leads to the fact that inland areas have vast areas, complex plateaus, and diverse climate types. It is necessary to construct T according to the terrain characteristics of each region. m Model.

[0005] The existing atmospheric weighted mean temperature fitting models are mainly divided into two categories: one is the T mThere are two kinds of experience models, one is to use the sounding data and surface temperature (T s ) data in North America, and find that the T m -T s linear regression bevis model (T m =0.72T s +70.2) is suitable for North America; one is obtained by linear regression by Zofia Baldysz et al using data from 109 sounding stations in Europe, and the T m model suitable for Europe is obtained; another kind is to construct the T m model suitable for the region according to the characteristics of the region, but the Yunnan-Guizhou Plateau region has rugged terrain, large altitude gradient and complex climate conditions, so that the commonly used T m empirical model has low prediction accuracy and large difference from the weighted mean temperature of the region, and cannot provide reliable meteorological information in time, therefore, a new method for constructing a regional weighted mean temperature model considering the product of multiple factors is proposed. SUMMARY

[0006] The purpose of the present application is to provide a new method for constructing a regional weighted mean temperature model considering the product of multiple factors, which solves the problems in the above background art.

[0007] To achieve the above purpose, the present application provides the following technical scheme: a new method for constructing a regional weighted mean temperature model considering the product of multiple factors, comprising the following steps:

[0008] S1, using the radio sounding station observation data in the Yunnan-Guizhou Plateau region, first, the data is preprocessed to eliminate obvious outliers, process missing values, and ensure the consistency of data formats of different sounding stations by vertical layering standardization of pressure type data;

[0009] S2, using the sounding station observation data and surface data in the region, the accurate atmospheric weighted mean temperature is calculated by numerical integration method, in order to screen the key factors for modeling, analyze the linear correlation between the atmospheric weighted mean temperature and the surface temperature, latitude, elevation, etc., and provide correlation factors for subsequent modeling according to the strength of correlation;

[0010] S3, based on the correlation analysis result, using the sounding station data in the region, a nonlinear regression model considering the product of multiple factors is constructed.

[0011] Preferably, the numerical integration method is used to calculate T m , and the nonlinear least squares method is used to construct the model as follows:

[0012] The integral of the corresponding water vapor pressure e and absolute temperature T in the atmospheric top layer direction is obtained, T mThe specific numerical integration method of the equation (1) is expressed as equation (2), which is expressed as:

[0013]

[0014] Wherein: T represents the absolute temperature (in Kelvin), e represents the water vapor pressure (in hPa), z represents the geopotential height in the zenith direction, and the radiosonde observation provides the data of relative humidity (RH) and absolute temperature T (unit: K), so that e can be calculated.

[0015] Preferably, the calculation equation of e is expressed as equation (2):

[0016]

[0017] In equation (2), Td represents the atmospheric temperature in Celsius (T = Td + 273.15), and in practical application, equation (1) is discretized to obtain equation (3), which is expressed as:

[0018]

[0019] In equation (3), ΔZi represents the thickness of the i-th atmospheric layer (m); n represents the number of atmospheric layers; e i and T i respectively represent the average temperature and water vapor pressure of the i-th atmospheric layer.

[0020] Preferably, the T m The main factors of change are daily cycle, annual cycle, semi-annual cycle, surface temperature, water vapor pressure, latitude and elevation. Considering the above influencing factors, the YGTm model is constructed to obtain equation (4), which is expressed as:

[0021] YGTm = ∫ (DOY, UT, T s , e s , Latitude, H) = ∫1·∫2·∫3 (4).

[0022] Preferably, the calculation method of ∫1 obtains equation (5), which is expressed as:

[0023]

[0024] Preferably, the calculation method of ∫2 obtains equation (6), which is expressed as:

[0025]

[0026] Preferably, the calculation method of ∫3 obtains equation (7), which is expressed as:

[0027] ∫3=e+∫·T S +g·lne s +h·Latitude+i·H (7).

[0028] Preferably, construct the T m When modeling, the influence of temperature changes during the day must be considered. Therefore, in equation (5), ∫1 represents T m where UT stands for Universal Time, and a1 and b1 are the coefficients of equation (5).

[0029] Preferably, the T m It has obvious periodic changes, namely interannual and semiannual changes. m This is taken into account when modeling. In equation (6), ∫2 represents T m The annual and semiannual change components of DOY show the day of the year, c i and d i is the coefficient, i=1, 2.

[0030] Preferably, in said S2, the Pearson correlation coefficient R is used to analyze T m The correlation between the factors and the influence of the relationship is as follows: m When modeling, we need to consider T S 、lne s , latitude and altitude m The influence of model accuracy on the T S 、lne s , latitude and elevation are considered as influencing factors and condensed in equation (7), where the units of latitude and elevation (H) are degrees and meters, while e, f, g, h and i are coefficients.

[0031] The present invention provides a new method for constructing a regional weighted average temperature model that takes into account the product of multiple factors. The new method for constructing a regional weighted average temperature model that takes into account the product of multiple factors has the following beneficial effects:

[0032] The present invention only needs to provide the surface meteorological values ​​at any time, and uses the model to calculate the atmospheric weighted average temperature, which is different from the commonly used T m The prediction model takes into account the special terrain requirements and climate change of the Yunnan-Guizhou Plateau, and also considers the influence of spatiotemporal dynamic characteristics, thus improving the prediction ability of the model and ensuring T m The high resolution and reliability of the calculation. For the technical effects, please refer to the corresponding calculation results in the specific implementation plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Fig. 1 A flow chart of the method for constructing the present invention;

[0034] Fig. 2 This is the distribution map of the sounding stations on the Yunnan-Guizhou Plateau;

[0035] Fig. 3 There are four types of T m Daily bias and RMS value of the model in different seasons. DETAILED DESCRIPTION

[0036] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described with reference to the accompanying drawings.

[0037] Example 1

[0038] A preferred embodiment of a new method for constructing a regional weighted average temperature model taking into account the product of multiple factors provided by the present invention is as follows: Figs. 1 to 3 As shown: A new method for constructing a regional weighted average temperature model taking into account the product of multiple factors includes the following steps:

[0039] S1. Using the radiosonde station location data in the Yunnan-Guizhou Plateau, select the radiosonde stations for modeling and the radiosonde stations for non-modeling.

[0040] Obtain the location information of all sounding stations in the study area, including the station number, longitude, latitude, altitude, etc., and display the location of the sounding stations in the area on the map. Leave the non-modeling stations in the five directions of up, down, left, right and center for subsequent model accuracy assessment. The station selection is as follows: Fig. 2 shown.

[0041] S2. Using radiosonde observation data from the Yunnan-Guizhou Plateau, we first preprocessed the data to remove obvious outliers and handle missing values. We then normalized the pressure data by vertical stratification to ensure that the data formats of different radiosonde stations were consistent.

[0042] Data from all sounding stations in the study area were obtained, including sounding observations multiple times a day (usually twice a day, at 00:00 UTC and 12:00 UTC), covering pressure level data (temperature, dew point temperature, relative humidity, air pressure, geopotential height) and surface data (surface temperature, water vapor pressure, station latitude and longitude, elevation). The data preprocessing steps are as follows: Quality check: Obvious outliers (such as temperature below -100°C or above 50°C, relative humidity exceeding 100%, etc.) were eliminated. Missing value processing: For a small amount of missing data, linear interpolation or sliding average of the previous and next time series was used to fill in the missing data; if the missing data exceeded 3 consecutive hours, it was marked as invalid data. Data standardization: The pressure level data were vertically layered and standardized to ensure that the data format of different sounding stations was consistent.

[0043] S3. Using the observation data from the radiosonde stations and the surface data in the region, calculate the accurate atmospheric weighted mean temperature by numerical integration method.

[0044] The integral of the corresponding water vapor pressure e and absolute temperature T at the radiosonde station toward the top of the atmosphere is used to obtain T m The specific numerical integration method is expressed by equation (1):

[0045]

[0046] where T represents the absolute temperature (in Kelvin), e represents the water vapor pressure (in hPa), and z represents the geopotential height in the zenith direction. Radiosonde observations provide data on relative humidity (RH) and absolute temperature T (in K), which enables the calculation of e.

[0047]

[0048] In equation (2), Td represents the atmospheric temperature in degrees Celsius (T=Td+273.15). In practical applications, equation (1) is discretized to obtain equation (3), which is expressed as:

[0049]

[0050] In equation (3), ΔZi represents the thickness of the i-th atmospheric layer (m); n represents the number of atmospheric layers; e i and T i represent the average temperature and water vapor pressure of the i-th atmospheric layer, respectively.

[0051] S4. Analyze the linear correlation between the atmospheric weighted mean temperature and the surface temperature, latitude, elevation, etc., and provide a correlation factor for subsequent modeling based on the strength of the correlation;

[0052] In addition, to study the region T m In this paper, the Pearson correlation coefficient R was used to analyze the correlation between T and the influencing factors using the measured data of 16 radiosonde stations in the Yunnan-Guizhou Plateau from 2010 to 2018. m With T S 、lne s , latitude and elevation. When R is greater than 0.5, it is highly correlated, between 0.3 and 0.5, it is moderately correlated, and between 0.1 and 0.3, it is weakly correlated. m With T S and lne s The correlation coefficients between them were 0.89 and 0.84, indicating that there was a strong correlation. m The correlation coefficients with elevation and latitude are 0.34 and 0.39, respectively, showing moderate correlation.

[0053] S5. Based on the results of the correlation analysis, a nonlinear regression model was constructed using the data from the radiosonde stations in the region, taking into account the spatiotemporal dynamics and the product of multiple factors.

[0054] This method studies the influence of T m The main factors of change are diurnal cycle, annual cycle, semiannual cycle, surface temperature, water vapor pressure, latitude and elevation. Taking into account the above influencing factors, the YGTm model is constructed as shown in equation (4);

[0055] YGTm=∫(DOY,UT,T s ,e s ,Latitude,H)=∫1·∫2·∫3 (4);

[0056]

[0057] ∫3=e+∫·T S +g·lne s +h·Latitude+i·H (7);

[0058] In building T m When modeling, the effect of temperature changes during the day must be considered. Therefore, in equation (5), ∫1 represents T m where UT stands for Universal Time, and a1 and b1 are the coefficients of equation (5).

[0059] The results of the study showed that T m It has obvious periodic changes (i.e., interannual and semiannual variations). m This is taken into account when modeling. As shown in Equation (6), ∫2 represents T m The annual and semiannual components of the change, DOY shows the day of the year, c i and d i is the coefficient, i=1, 2.

[0060] In addition, the Pearson correlation coefficient R was used in step 4 to analyze T m The correlation between the factors. m When modeling, we need to consider T s 、lne s , latitude and altitude m Therefore, the T s 、lne s , latitude, and elevation are considered as influencing factors and condensed in equation (7). The units of latitude and elevation (H) are degrees and meters, while e, f, g, h, and i are coefficients.

[0061] Substituting the sounding data of the 16 modeling stations in this area into equation (4) yields the corresponding coefficients, as shown in Table 1:

[0062]

[0063]

[0064] In addition, the corresponding sounding station data are brought into the YGTm model to obtain the corresponding atmospheric weighted average temperature.

[0065] Example 2

[0066] The present invention provides a preferred embodiment of a new method for constructing a regional weighted average temperature model taking into account the product of multiple factors. The atmospheric weighted average temperature model of this technical solution is compared with the prior art:

[0067] 1. Bevis model (T m =70.2+0.72T s );

[0068] 2. HGPT2 model: Global elevation correction model

[0069] 3. NTm model: A model for western China based on ground temperature, altitude, and seasonal changes:

[0070]

[0071] This analysis uses the T calculated from the 2019 modeling sounding station data in the Yunnan-Guizhou Plateau. m The data is used as a test sample, and the average deviation (Bias) and root mean square error (RMS) are used to evaluate the modeling accuracy. The accuracy comparison of the above models is shown in Table 2:

[0072]

[0073] Through comprehensive analysis of the data in Table 2, we can get the following: the maximum deviation value of the YGTm model is 1.30K, the minimum deviation value is -2.25K and the average deviation value is -0.17K. It is worth noting that, compared with other T m Compared with the model, the T calculated by YGTm model m No obvious systematic deviation is shown. The average RMS value of the YGTm model is 2.42 K, which is reduced by 28.8% (0.98 K) compared with the Bevis model. In addition, it reduces the average RMS by 20.7% (0.63 K) and 30.7% (1.07 K) compared with the HGPT2 model and the NTm model, respectively. In summary, the YGTm model has a good performance in the selected T m The model has the highest prediction accuracy for the Yunnan-Guizhou Plateau.

[0074] At the same time, these T m The daily deviations and RMS values ​​of the models were used to evaluate the seasonal performance of the four selected models, as Fig. 3 As shown. The Bevis, HGPT2, and NTm models exhibit significant seasonal and systematic biases, which are particularly evident in spring and winter. However, the YGTm model shows smaller bias values ​​and no significant seasonal variations throughout the year. Regarding the RMS values, all models show significant seasonal variations, with larger RMS values ​​in spring and winter and smaller RMS values ​​in summer. This pattern can be attributed to the fact that most of the selected radiosonde stations are located in mid-latitudes, where T m The changes are smaller in summer and larger in winter.

[0075] This analysis uses the T calculated from the non-modeling radiosonde station data in the Yunnan-Guizhou Plateau in 2019. m The data is used as the test sample, and the average deviation (Bias) and root mean square error (RMS) are used to evaluate the modeling accuracy. The accuracy comparison of the above models is shown in Table 3:

[0076]

[0077]

[0078] In Table 3, the Bevis, HGPT2, and NTm models show significant negative biases. This difference highlights the obvious systematic biases in the calculations of the Bevis, NTm, and HGPT2 models. Compared with the other three models, the average bias of the YGTm model is significantly lower, at -0.30 K. The high accuracy of the YGTm model is due to the T m Regarding the RMS value, the average RMS value of the YGTm model is 2.61K. Compared with the Bevis, NTm and HGPT2 models, the accuracy of the YGTm model (in terms of RMS value) is improved by 31.7% (1.21K), 26.2% (0.93K) and 23.2% (0.79K). In summary, compared with the other three selected T m Compared with the three models, the YGTm model showed better prediction accuracy and stability.

[0079] Example 3

[0080] The formula for GNSS-derived atmospheric water vapor content is PWV = ZWD·Π, where ZWD represents the zenith wet delay; PWV is the atmospheric water vapor content; and Π is the water vapor conversion coefficient. The calculation formula is:

[0081]

[0082] In the above formula: ρw and R v denote the density of liquid water and the specific gas constant of water vapor, respectively; k2' and k3 denote the atmospheric refractive index constants; in this case, we should focus on improving T m The prediction accuracy of . Following the error propagation law, an approximate formula is derived by differentiation:

[0083]

[0084] because Here k'2=17±10K / mbar,k3=3.776±0.014×10 5 K 2 / mbar, while in the present invention T m In the range of 245-305K, the above formula can be simplified to the following equation:

[0085]

[0086] Based on the above equation, PWV and T m The relationship between RMS PWV represents the RMS value of PWV; T m RMS value. When T is introduced m When the average annual number of PWV and RMS is used, the relative error of PWV can be estimated using the equation. PWV and RMS PWV PWV is used to assess T m The impact of errors on the GNSS PWV it generates.

[0087] The same method was used to analyze different m The impact of model values ​​on GNSS PWV. The results are as follows

[0088] As shown in Table 4:

[0089]

[0090] In Table 4, the RMS associated with the YGTm model PWV The value is always lower than 0.39mm, and the average RMS PWV The value is 0.26mm. About RMS PWV / PWV, the average value of the YGTm model was 0.87%, ranging from 0.64% to 1.16%. The YGTm model provides accurate T m The value is enhanced, which enhances its ability to accurately invert PWV over the Yunnan-Guizhou Plateau.

[0091] The above description is only an illustrative embodiment of the present invention and is not intended to limit the scope of the present invention. Any equivalent changes and modifications made by any person skilled in the art without departing from the concept and principle of the present invention should fall within the scope of protection of the present invention. It should also be noted that the various components of the present invention are not limited to the above-mentioned overall application. The various technical features described in the specification of the present invention can be selected one by one or multiple by multiple combinations according to actual needs. Therefore, the present invention should naturally cover other combinations and specific applications related to this case.

Claims

1. A new method for constructing a regional weighted average temperature model taking into account the product of multiple factors, characterized in that: The following steps are involved: S1. Using radiosonde observation data from the Yunnan-Guizhou Plateau, we first preprocessed the data to remove obvious outliers and handle missing values. We then normalized the pressure data by vertical stratification to ensure that the data formats of different radiosonde stations were consistent. S2. Using the observation data from the radiosonde stations and the surface data in the region, the precise atmospheric weighted average temperature is calculated by numerical integration. To screen the key factors for modeling, the linear correlation between the atmospheric weighted average temperature and the surface temperature, latitude, and elevation is analyzed. The correlation factor is provided for subsequent modeling based on the strength of the correlation. S3. Based on the results of correlation analysis, a nonlinear regression model taking into account the product of multiple factors was constructed using the sounding station data in the region.

2. A new method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 1, characterized in that: The numerical integration method is used to calculate T m The nonlinear least squares method is used to construct the model as follows: The integral of the corresponding water vapor pressure e and absolute temperature T at the radiosonde station toward the top of the atmosphere is used to obtain T m The specific numerical integration method is expressed as equation (1), which is expressed as: where T represents the absolute temperature (in Kelvin), e represents the water vapor pressure (in hPa), and z represents the geopotential height in the zenith direction. Radiosonde observations provide data on relative humidity (RH) and absolute temperature T (in K), which enables the calculation of e.

3. The novel method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 2, characterized in that: The calculation equation of e is expressed as (2): In equation (2), Td represents the atmospheric temperature in degrees Celsius (T=Td+273.15). In practical applications, equation (1) is discretized to obtain equation (3), which is expressed as: In equation (3), ΔZi represents the thickness of the i-th atmospheric layer (m); n represents the number of atmospheric layers; e i and T i represent the average temperature and water vapor pressure of the i-th atmospheric layer, respectively.

4. The novel method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 2, characterized in that: The T m The main factors of change are diurnal cycle, annual cycle, semi-annual cycle, surface temperature, water vapor pressure, latitude and elevation. Taking these factors into consideration, the YGTm model is constructed to obtain equation (4), which is expressed as: YGTm=∫(DOY,UT,T s ,e s ,Latitude,H)=∫1·∫2·∫3 (4)。 5. A new method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 4, characterized in that: The calculation method of ∫1 obtains equation (5), which is expressed as:

6. A new method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 4, characterized in that: The calculation method of ∫2 obtains equation (6), which is expressed as:

7. A new method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 4, characterized in that: The calculation method of ∫3 obtains equation (7), which is expressed as: ∫3=e+∫·T S +g·lne s +h·Latitude+i·H (7)。 8. The novel method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 2, characterized in that: Construct the T m When modeling, the influence of temperature changes during the day must be considered. Therefore, in equation (5), ∫1 represents T m where UT stands for Universal Time, and a1 and b1 are the coefficients of equation (5).

9. The novel method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 1, characterized in that: The T m It has obvious periodic changes, namely interannual and semiannual changes. m This is taken into account when modeling. In equation (6), ∫2 represents T m The annual and semiannual change components of DOY show the day of the year, c i and d i is the coefficient, i=1, 2.

10. The novel method for constructing a regional weighted average temperature model taking into account multi-factor products according to claim 1, characterized in that: In S2, Pearson correlation coefficient R was used to analyze T m The correlation between the factors and the influence of the relationship is as follows: m When modeling, we need to consider T S 、lne s , latitude and altitude m The influence of model accuracy on the T S 、lne s , latitude and elevation are considered as influencing factors and condensed in equation (7), where the units of latitude and elevation (H) are degrees and meters, while e, f, g, h and i are coefficients.

Images