Uniform hourly air temperature data processing method for regional automatic weather station

Abstract

The invention discloses a uniform hourly air temperature data processing method for regional automatic weather stations. The method comprises the following steps: collecting data sources (regional automatic weather station observation data and regional automatic weather station metadata) of each regional automatic weather station and national automatic weather station observation data; performing quality control on the observation data of the regional automatic weather station by adopting climate boundary value inspection, internal consistency inspection, inspection that the temperature value is negative from June to August, climate abnormal value inspection and space consistency inspection; screening a regional automatic weather station; carrying out homogenization analysis to obtain homogenization hourly air temperature data of the regional automatic weather station; performing data quality evaluation on the uniform hourly air temperature data of the regional automatic weather station; and analyzing air temperature trend characteristics. The invention relates to the technical field of meteorological observation data processing, and can solve the problems that in the prior art, a climate data set lacks a data source of a regional automatic meteorological station, a homogenization processing technology is not enough in application, and a research analysis result is uncertain.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological observation data processing, and in particular to a method for processing homogenized hourly temperature data of a regional automatic weather station. Background Art

[0002] The observation records of meteorological stations are the basic guarantee for the development of meteorological services and the effective supply of high-quality meteorological information in meteorological services. In particular, long-term continuous and reliable high-temporal and spatial resolution observation data are important basic support for revealing the occurrence and development of regional and local extreme climate events and their mechanisms. In recent years, my country's meteorological data quality control technology has made breakthrough progress, realizing the identification and correction of erroneous information in real-time observation data. However, in actual observations, due to station migration, instrument changes, etc., there are inevitably obvious non-uniformity problems in long-term observation sequences, resulting in a lack of reliable observation basis for detecting regional or local climate change laws and predicting future climate change trends. In addition, with the modernization and refinement of my country's meteorological services, the regional automatic weather station observation system has entered a mature stage. Compared with national meteorological observation stations, it has the advantage of dense site coverage and can comprehensively and systematically capture regional and local weather and climate change processes. However, at present, due to outstanding problems such as data integrity and time series homogeneity, regional automatic station observation data has not been well applied in weather, climate and extreme climate change services.

[0003] At present, the most representative long-term observational climate datasets in the world mainly include the Global Surface Temperature Dataset (GISTEMP) of the United States, the Global Historical Climate Dataset (GHCN-monthly / daily), the Berkeley Earth Surface Temperature Dataset (Berkeley Earth), and the Global Monthly Mean Surface Climate Elements Dataset (CRU) of the United Kingdom. These datasets are all based on daily or monthly observation data from national meteorological stations on all continents around the world. They are obtained by integrating data sources from different stations, quality control of temporal and spatial uniformity, and analysis of climate anomalies, but the inhomogeneity in the observation series has not been further processed. Chinese scholars have also conducted various studies on the establishment of long-term climate series, and based on the observation data after quality control, a series of temperature data products have been released in China. Cao et al. (2013) established a homogenized monthly mean temperature dataset for central and eastern China since the end of the 19th century. Yun et al. (2019) updated the global land surface temperature dataset (CMA-LSAT) established by Xue et al. (2018) to obtain an integrated global surface monthly temperature dataset (CMST). To further improve the coverage of global surface temperature data, Sun et al. (2021) constructed a 5°×5° global temperature dataset (CMST-Interim) based on the latest updated global land surface temperature dataset (C-LSAT 2.0) by Li et al. (2021). These datasets provide a relatively reliable observational basis for global and regional climate change in China, but they are also mainly based on observational data from national meteorological stations.

[0004] Chinese invention patent CN112559588B discloses a method for constructing a century-long homogenized daily temperature sequence observed by ground meteorological stations, and specifically discloses: establishing basic data of the century-long homogenized daily temperature sequence, splicing the data to form a complete century-long sequence, and then performing quality control on it, performing a homogenization test on the century-long sequence, establishing a monthly reference sequence in the homogenization test, and obtaining breakpoints through the test; then, under the daily value reference sequence, based on the correction with the reference sequence method, obtaining the century-long observed homogenized daily temperature sequence; for the century-long observed homogenized daily temperature sequence, obtaining the optimal interpolation sequence through data interpolation; evaluating the optimal interpolation sequence to obtain the final homogenized daily temperature sequence.

[0005] The quality control of this construction method uses a dynamic f-value test, which is only applicable to the processing of national automatic station observation data and requires quality control of at least 30 years of observation data; the internal consistency check of this construction method considers data beyond the threshold range as erroneous data, resulting in low reliability of missing measurement judgment; this construction method uses 5 times the standard deviation to check climate anomalies, resulting in low reliability and accuracy of the data.

[0006] The homogeneity test method of this construction method adopts the Penalized Maximum T Test (PMT), and the determination of breakpoints is divided into two cases: one is that in the absence of historical evolution data and archival information of ground meteorological stations, the significant sequence breakpoints that are tested in both reference sequences are selected; the other is that in the case of historical evolution data and archival information of ground meteorological stations, if the time position of the significant sequence breakpoint obtained by the test is consistent with the time recorded in these data, then the breakpoint is retained, and the quantile matching method is used to correct the homogeneity of these breakpoints. There are artificial random errors in the homogenization process, which increases the uncertainty of the research and analysis results.

[0007] In summary, the long-term observational climate data sets of the prior art have had a profound impact on the scientific development of global / regional climate and extreme climate change research, but most of the established data sets lack the data sources of regional automatic weather stations with more precise and dense temporal and spatial distribution. At the same time, in the established climate series, the application of homogenization processing technology is insufficient, which seriously hinders the in-depth study of the changes in long-term extreme climate events. Therefore, it is necessary to provide a method for processing homogenized hourly temperature data of regional automatic weather stations, which can solve the problems in the prior art that the climate data sets lack the data sources of regional automatic weather stations, the application of homogenization processing technology is insufficient, and there is uncertainty in the research and analysis results. Summary of the invention

[0008] The purpose of the present invention is to provide a method for processing homogenized hourly temperature data of a regional automatic weather station, which can solve the problems in the prior art that the climate data set lacks the data source of the regional automatic weather station, the application of the homogenization processing technology is insufficient, and there is uncertainty in the research and analysis results.

[0009] The present invention is achieved in that:

[0010] A method for processing homogenized hourly temperature data of a regional automatic weather station comprises the following steps:

[0011] Step 1: Collect data sources of automatic weather stations in each region and observation data of national automatic weather stations;

[0012] The data sources of regional automatic weather stations include regional automatic weather station observation data and regional automatic weather station metadata;

[0013] Step 2: Perform data quality control on the regional automatic weather station observation data collected in step 1, and use climate limit value check, internal consistency check, negative value check of June-August temperature, climate anomaly check and spatial consistency check to control the quality of the regional automatic weather station observation data;

[0014] Step 3: Effectively screen regional automatic weather stations based on data sources that have undergone quality control according to data integrity and time series continuity;

[0015] Step 4: Perform homogenization analysis on the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the selected regional automatic weather stations to obtain the homogenized hourly temperature data of the regional automatic weather stations;

[0016] Step 5: Conduct data quality assessment on the homogenized hourly temperature data of the regional automatic weather station;

[0017] Step 6: Analyze temperature trend characteristics.

[0018] The regional automatic weather station is an unmanned station; the observation data of the regional automatic weather station includes the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the regional automatic weather station since its establishment;

[0019] The metadata of the regional automatic weather station includes the time of relocation, element upgrade time, and collector update time of the regional automatic weather station since its establishment;

[0020] The national automatic weather station is a manned station, and the observation data of the national automatic weather station include the daily average temperature, daily maximum temperature and daily minimum temperature of the national automatic weather station after uniformity correction since its establishment.

[0021] In the step 2, the specific operation of the climate limit value check is: set the hourly average temperature outside the range of -80℃-60℃ as missing, set the hourly maximum temperature outside the range of -80℃-60℃ as missing, and set the hourly minimum temperature outside the range of -80℃-60℃ as missing.

[0022] In the step 2, the specific operation of the internal consistency check is: set the judgment logic: hourly minimum temperature ≤ hourly average temperature ≤ hourly maximum temperature. For the regional automatic weather station observation data that does not meet the judgment logic, if the regional automatic weather station observation data is reasonable in its own time series changes, the hourly maximum temperature or the hourly minimum temperature is modified based on the hourly average temperature. However, if it is manually verified to be erroneous data, it is set to missing.

[0023] In the step 2, the specific operation of checking whether the temperature values ​​from June to August are negative is: when the observation data of the regional automatic weather station from June to August are negative, the observation data of the regional automatic weather station are set as missing.

[0024] In step 2, the specific operation of checking the climate anomaly is: using formula (1), comparing the observation data of the regional automatic weather station with the corresponding daily standard deviation temperature value, and when it is manually verified to be erroneous data, setting the observation data of the regional automatic weather station as missing;

[0025]

[0026] Among them, i is a day, j is an hour on a day, and x is ij The hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked. is the average value of the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked, σ i is the standard deviation of the daily sequence of hourly average temperature, hourly maximum temperature or hourly minimum temperature of the regional automatic weather station being checked, n is the multiple of standard deviation, n = 3-5;

[0027] If the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the day is missing for 6 times or more, the statistical value of the daily series mean or standard deviation will be regarded as missing and no statistical check will be performed.

[0028] In step 2, the specific operation of the spatial consistency check is: using formula (2), the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is compared and analyzed with the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the adjacent area respectively. If it is manually verified to be erroneous data, the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is set as missing;

[0029] |Z t -Z t0 |>nσ t0 , (2)

[0030] Among them, Z t and Z t0 are the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the standardized series of the automatic weather station in the inspected area and the automatic weather station in the adjacent area, respectively; t0 The standard deviation of daily values ​​of the normalized series of the neighboring regional automatic weather stations constructed by arithmetic averaging.

[0031] The step 3 comprises the following sub-steps:

[0032] Step 3.1: Select the regional automatic weather stations whose observation series of hourly average temperature, hourly maximum temperature and hourly minimum temperature are at least 10 years long and have continuous and valid observation data during this period;

[0033] Step 3.2: Among the regional automatic weather stations selected in step 3.1, further select regional automatic weather stations whose hourly average temperature, hourly maximum temperature and hourly minimum temperature meet the selection conditions of step 3.1.

[0034] The step 4 comprises the following sub-steps:

[0035] Step 4.1: Count the hourly average temperature, hourly maximum temperature, and hourly minimum temperature data into daily average temperature, daily maximum temperature, and daily minimum temperature, respectively. Use the maximum penalty F test to test the homogeneity of the daily average temperature, daily maximum temperature, and daily minimum temperature series, with a significance level of α of 0.01, to obtain the breakpoints of the series heterogeneity.

[0036] Step 4.2: Verify the authenticity of the sequence heterogeneity breakpoints. Based on the metadata of the regional automatic weather station, remove the statistically significant sequence heterogeneity breakpoints that are not supported by the exact regional automatic weather station metadata, retain the statistically significant sequence heterogeneity breakpoints that are supported by the exact regional automatic weather station metadata, and determine the regional automatic weather stations with such characteristics as heterogeneous regional automatic weather stations;

[0037] Step 4.3: Eliminate regional automatic weather stations with non-uniform characteristics to form the homogenized hourly temperature data of regional automatic weather stations.

[0038] The step 5 comprises the following sub-steps:

[0039] Step 5.1: Count the hourly average temperature, hourly maximum temperature and hourly minimum temperature after homogenization analysis in step 4 as daily values;

[0040] Step 5.2: Perform data quality assessment on the average sequence of the regional automatic weather station corresponding to the daily hourly average temperature, hourly maximum temperature and hourly minimum temperature. The assessment index is the standard deviation of the difference sequence of observation data between the regional automatic weather station and the national automatic weather station.

[0041] Step 5.3: Repeat steps 5.1 and 5.2 to evaluate the data quality of the automatic weather stations in each region after screening.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention takes regional automatic weather stations as the research object, and constructs a true and reliable long-term homogenized hourly temperature observation sequence through data quality control and homogenization analysis and processing, thereby improving the quality of long-sequence hourly observation data of regional automatic weather stations, and enhancing their application value in the service support capabilities for responding to and adapting to climate change, providing a newer and more precise observation basis for the construction of my country's service support capabilities for responding to extreme climate change, and processing and analyzing the observation data of regional automatic weather stations more quickly and effectively, simplifying the processing steps and improving the application efficiency of data.

[0044] 2. The data quality control of the present invention adopts negative temperature values ​​from June to August and spatial consistency check according to the distribution characteristics of the observation data of the regional automatic weather station and the length of the observation sequence, so that it can be applicable to the processing of the observation data of the regional automatic weather station, and performs internal consistency check through logical judgment and manual verification of erroneous information. The setting of 3-5 times the standard deviation reduces the irreversible influence of human factors on the reliability of first-hand observation data, thereby improving the reliability and correctness of the analysis results of the observation data of the regional automatic weather station.

[0045] 3. The homogenization analysis of the present invention is based on the reference-free sequence method. Through the penalized maximum F test, it is determined that the statistically significant breakpoints with definite regional automatic weather station metadata support are non-uniform regional automatic weather stations. There is no need to perform uniformity correction. The non-uniform regional automatic weather stations are directly deleted, which can effectively reduce the random errors introduced by human processing to the original observation data and ensure the certainty of the research and analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flow chart of a method for processing homogenized hourly temperature data of a regional automatic weather station. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0048] Please see attached Figure 1 , a method for processing homogenized hourly temperature data of a regional automatic weather station, comprising the following steps:

[0049] Step 1: Collect data sources of automatic weather stations in each region and observation data of national automatic weather stations.

[0050] The data source of the regional automatic weather station includes regional automatic weather station observation data and regional automatic weather station metadata.

[0051] The regional automatic weather station is an unmanned station; the observation data of the regional automatic weather station include the hourly average temperature, hourly maximum temperature and hourly minimum temperature since the establishment of the regional automatic weather station.

[0052] The metadata of the regional automatic weather station includes information such as the time of station relocation, element upgrade time, collector update time, etc. since the establishment of the regional automatic weather station.

[0053] The national automatic weather station is a manned station, and the observation data of the national automatic weather station include the daily average temperature, daily maximum temperature and daily minimum temperature of the national automatic weather station after uniformity correction since its establishment.

[0054] Step 2: Perform data quality control on the regional automatic weather station observation data collected in step 1, using climate limit value check, internal consistency check, check for negative temperature values ​​from June to August, climate anomaly check and spatial consistency check to control the quality of the regional automatic weather station observation data.

[0055] In the step 2, the specific operation of the climate limit value check is: set the hourly average temperature outside the range of -80℃-60℃ as missing, set the hourly maximum temperature outside the range of -80℃-60℃ as missing, and set the hourly minimum temperature outside the range of -80℃-60℃ as missing.

[0056] In the step 2, the specific operation of the internal consistency check is: set the judgment logic: hourly minimum temperature ≤ hourly average temperature ≤ hourly maximum temperature. For the regional automatic weather station observation data that does not meet the judgment logic, if the regional automatic weather station observation data is reasonable in its own time series changes, the hourly maximum temperature or the hourly minimum temperature is modified based on the hourly average temperature. However, if it is manually verified to be erroneous data, it is set to missing.

[0057] The hourly average temperature, hourly maximum temperature and hourly minimum temperature all conform to the time variation law of the observation series. In the process of climate time series change, no significant abnormal values ​​appear. At this time, if there is a discrepancy with the set judgment logic, the hourly maximum temperature or hourly minimum temperature will be modified based on the hourly average temperature.

[0058] Specifically, the hourly minimum temperature, hourly average temperature and hourly maximum temperature are 26.5℃, 25.0℃ and 25.5℃ respectively. Based on the hourly average temperature, the hourly minimum temperature will be modified to 25.0℃. The hourly minimum temperature, hourly average temperature and hourly maximum temperature are 14.5℃, 17.0℃ and 15.5℃ respectively. Based on the hourly average temperature, the hourly maximum temperature will be modified to 17.0℃.

[0059] For regional automatic weather station observation data that does not meet the judgment logic, that is, the suspected erroneous information found during the inspection, each piece of suspected erroneous information is manually verified, including judging the authenticity of the suspected erroneous information by consulting the metadata of the corresponding regional automatic weather station, and manually analyzing the logical evolution of the temperature time series, etc., to ensure the accuracy of data quality control.

[0060] In the step 2, the specific operation of checking whether the temperature values ​​from June to August are negative is: when the observation data of the regional automatic weather station from June to August are negative, the observation data of the regional automatic weather station are set as missing.

[0061] In step 2, the specific operation of checking the climate anomaly is: using formula (1), comparing the observation data of the regional automatic weather station with the corresponding daily standard deviation temperature value, and when it is manually verified to be erroneous data, setting the observation data of the regional automatic weather station as missing;

[0062]

[0063] Among them, i is a day, j is an hour on a day, and x is ij The hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked. is the average value of the daily sequence of hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the checked area, σ i is the standard deviation of the daily sequence of the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the regional automatic weather station being checked, n is the multiple of the standard deviation, preferably, n=3-5.

[0064] If the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the day is missing for 6 times or more, the statistical value of the daily series mean or standard deviation will be regarded as missing and no statistical check will be performed.

[0065] By setting the standard deviation multiple n and not conducting statistical checks on hourly average temperatures that are missing six or more times, the irreversible impact of human factors on the reliability of first-hand observation data can be effectively reduced, thereby improving the reliability and accuracy of the data.

[0066] In step 2, the specific operation of the spatial consistency check is: using formula (2), the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is compared and analyzed with the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the adjacent area respectively. If it is manually verified to be erroneous data, the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is set as missing;

[0067] |Z t -Z t0 |>nσ t0 , (2)

[0068] Among them, Z t and Z t0 are the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the standardized series of the automatic weather station in the inspected area and the automatic weather station in the adjacent area, respectively; t0It is the daily standard deviation of the standardized series of the automatic weather stations in the neighboring areas of the regional automatic weather station constructed by arithmetic averaging. Its data source is the observation data of the regional automatic weather station. The statistical principle of the daily series standard deviation is consistent with the above-mentioned climate anomaly check step and will not be repeated here.

[0069] In order to obtain relatively reasonable inspection results, the standard deviation multiple n = 3-5 was proposed in the study, and the selection principle of automatic weather stations in the neighboring area was: 3, 5, and 20 stations that were spatially close to the inspected station, and constructed through arithmetic averaging.

[0070] By checking the negative values ​​of the temperature from June to August and checking the spatial consistency, we can better target the distribution characteristics of the observation data of the regional automatic weather station and the length of the observation sequence, which is suitable for the data processing of the regional automatic weather station.

[0071] Step 3: Effectively screen regional automatic weather stations based on data sources that have undergone quality control according to data integrity and time series continuity.

[0072] The step 3 comprises the following sub-steps:

[0073] Step 3.1: Select the regional automatic weather stations whose observation series of hourly average temperature, hourly maximum temperature and hourly minimum temperature are at least 10 years long and have continuous (10 years or more) valid observation data during this period.

[0074] Step 3.2: Among the regional automatic weather stations selected in step 3.1, further select regional automatic weather stations whose hourly average temperature, hourly maximum temperature and hourly minimum temperature meet the selection conditions of step 3.1.

[0075] Screening regional automatic weather stations can screen out regional automatic weather stations that are of significant significance to the field of climate time scale research, that is, regional automatic weather stations with sufficiently long time series and continuous and valid observation data, thereby ensuring the integrity and validity of the data.

[0076] Step 4: Perform homogenization analysis on the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the screened regional automatic weather stations to obtain the homogenized hourly temperature data of the regional automatic weather stations.

[0077] In the homogenization analysis process, in order to avoid the autocorrelation problem of the hourly observation sequence and obtain more reasonable statistically significant breakpoints, namely, the breakpoints of sequence heterogeneity, the hourly average temperature, hourly maximum temperature, and hourly minimum temperature are all counted as daily values. The reference-free sequence method is used to perform homogenization analysis through the Penalized Maximum F Test (PMFT), and the daily value test results are mainly used to verify the regional automatic weather station metadata (station relocation, element upgrade, collector update, etc.), so as to determine the sequence homogeneity of the regional automatic weather station observation data. Among them, the method for determining the breakpoint of sequence heterogeneity is to detect the breakpoint of sequence heterogeneity that is statistically significant (significance level α = 0.01) and supported by the metadata of the regional automatic weather station.

[0078] The step 4 comprises the following sub-steps:

[0079] Step 4.1: Count the hourly average temperature, hourly maximum temperature, and hourly minimum temperature data into daily average temperature, daily maximum temperature, and daily minimum temperature, respectively. Use the Penalized Maximum F Test (PMFT) to perform a homogeneity test on the daily average temperature, daily maximum temperature, and daily minimum temperature series, with a significance level of α of 0.01, to obtain the heterogeneity breakpoints of the series.

[0080] The penalized maximum F test (PMFT) is a conventional data processing method (e.g. Si et al., 2021; Si Peng et al., 2022; Yang Su, 2023, etc.), and the process of homogeneity test will not be described here. The penalized maximum F test (PMFT) has a good test effect on the breakpoints of sequence heterogeneity in the absence of a relatively reliable reference sequence (e.g. Cao et al., 2017; Si et al., 2024). Finally, by eliminating the regional automatic weather stations corresponding to the breakpoints of sequence heterogeneity that are statistically significant (significance level α = 0.01) and supported by the metadata of the exact regional automatic weather stations, a relatively uniform hourly temperature dataset of regional automatic weather stations in Tianjin is formed.

[0081] Step 4.2: Verify the authenticity of the sequence heterogeneity breakpoints. Based on the metadata of the regional automatic weather stations, eliminate the statistically significant sequence heterogeneity breakpoints that are not supported by the exact regional automatic weather station metadata, retain the statistically significant sequence heterogeneity breakpoints that are supported by the exact regional automatic weather station metadata, and determine the regional automatic weather stations with such characteristics as heterogeneous regional automatic weather stations.

[0082] Step 4.3: Eliminate regional automatic weather stations with non-uniform characteristics to form the homogenized hourly temperature data of regional automatic weather stations.

[0083] The observation data of regional automatic weather stations are characterized by dense spatial distribution and short observation time series. The present invention adopts the reference-free sequence method, and the homogeneity test method is the penalized maximum F test (PMFT). The determination of the sequence heterogeneity breakpoints is statistically significant and has definite sequence heterogeneity breakpoints supported by metadata of regional automatic weather stations. There is no need to perform homogeneity correction, but directly delete the regional automatic weather stations with sequence heterogeneity breakpoints, which can effectively reduce the random errors introduced by human processing to the original observation data and cause uncertainty in the results of climate research analysis.

[0084] Step 5: Conduct data quality assessment on the homogenized hourly temperature data of the regional automatic weather station.

[0085] The step 5 comprises the following sub-steps:

[0086] Step 5.1: Count the hourly average temperature, hourly maximum temperature and hourly minimum temperature after the homogenization analysis in step 4 as daily values.

[0087] Step 5.2: Perform data quality assessment on the average sequence of the regional automatic weather station corresponding to the daily hourly average temperature, hourly maximum temperature and hourly minimum temperature. The assessment indicator is the standard deviation of the difference sequence of observation data between the regional automatic weather station and the national automatic weather station.

[0088] Step 5.3: Repeat steps 5.1 and 5.2 to evaluate the data quality of the automatic weather stations in each region after screening.

[0089] Step 6: Analyze temperature trend characteristics.

[0090] Preferably, robust regression can be used to fit and analyze the trend variation amplitude and characteristics of the annual and seasonal series of the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the regional automatic weather station.

[0091] Embodiment 1: Taking Tianjin as an example, the observation data of the automatic weather station in the region are processed.

[0092] Step 1: The regional automatic weather station observation data used in this embodiment is derived from the Tianjin Meteorological Big Data Cloud Platform (Tianqing Database), and the hourly average temperature, hourly maximum temperature and hourly minimum temperature data (all in universal time) of 325 regional automatic weather stations in Tianjin from 2008 to 2023 are selected. At the same time, according to the metadata table of provincial meteorological observation station business access published by the Tianjin Meteorological Bureau over the years, the metadata information such as the station migration time, element upgrade time, collector update time, etc. of the 325 regional automatic weather stations since their establishment are sorted out as an objective basis for judging the homogeneity of the temperature time series. In addition, as reference materials, the daily average temperature, daily maximum temperature and daily minimum temperature data of the national meteorological stations in 11 administrative regions of Tianjin from 2012 to 2023 are derived from the homogenized daily temperature data set of Tianjin from 1951 to the present developed by Si Peng et al. (2015), which is used to evaluate and analyze the reliability of the homogenized hourly temperature data of regional automatic weather stations.

[0093] Step 2: Check climate limit values: Set the hourly average temperature, hourly maximum temperature, or hourly minimum temperature that is outside the range of -80℃ to 60℃ as missing.

[0094] Internal consistency check: Set the judgment logic: hourly minimum temperature ≤ hourly average temperature ≤ hourly maximum temperature. For regional automatic weather station observation data that do not meet the judgment logic, if the regional automatic weather station observation data is reasonable in its own time series changes, the hourly maximum temperature or hourly minimum temperature is modified based on the hourly average temperature value. However, if it is confirmed to be erroneous data after manual verification, it will be set as missing.

[0095] The results of internal consistency check show that 148 suspected errors were found in 15 regional automatic weather stations when the data of 325 stations were not missing. Among them, the suspected errors accounted for 68.2% in 2009-2010, mainly in three regional automatic weather stations, A2364 (Ninghe District), A3171 (Binhai New Area) and A3585 (urban area); 28.4% in 2012-2013, mainly in two regional automatic weather stations, A2169 (Jizhou District) and A2565 (Jinghai District). In the process of processing, the hourly maximum temperature or hourly minimum temperature was modified based on the hourly average temperature for the regional automatic weather stations A2169 and A3171; the hourly average temperature of the suspected errors in the regional automatic weather station A3585 was set as missing, and the hourly maximum temperature and hourly minimum temperature of the suspected errors in the regional automatic weather stations A2364 and A2565 were set as missing. Table 1 gives the revised results of the internal consistency check using the other 10 stations as examples.

[0096] Table 1 Correction results of internal consistency erroneous information

[0097]

[0098] Check for negative temperature values ​​from June to August: When the observation data of the regional automatic weather station from June to August is negative, the observation data of the regional automatic weather station will be set as missing.

[0099] Climate anomaly check: Use formula (1) to compare the observation data of the regional automatic weather station with its corresponding daily standard deviation temperature value. If it is confirmed to be erroneous data after manual verification, the observation data of the regional automatic weather station will be set as missing;

[0100]

[0101] Among them, i is a day, j is an hour on a day, and x is ij The hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked. is the average value of the daily sequence of hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the checked area, σ i is the standard deviation of the daily sequence of the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the regional automatic weather station being checked, n is the multiple of the standard deviation, preferably, n=3-5.

[0102] If the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the day is missing for 6 times or more, the statistical value of the daily series mean or standard deviation will be regarded as missing and no statistical check will be performed.

[0103] By setting the standard deviation multiple n and not conducting statistical checks on the hourly average temperature, hourly maximum temperature or hourly minimum temperature that are missing 6 times or more, the irreversible impact of human factors on the reliability of first-hand observation data can be effectively reduced, and the reliability and accuracy of the data can be improved.

[0104] The results of climate anomaly inspection showed that in the case of non-missing data, the hourly average temperature, hourly maximum temperature, and hourly minimum temperature of 325 regional automatic weather stations did not detect any suspicious information at the 5-times standard deviation threshold, and the 3-times standard deviation threshold detected a large amount of suspicious information, with 9431, 8722, and 7334 pieces of suspicious information for the hourly average temperature, hourly maximum temperature, and hourly minimum temperature, respectively. After preliminary manual verification, most of the suspicious information was credible, and the inspection results of the 4-times standard deviation threshold were relatively reasonable. Therefore, the final standard deviation multiple n = 4 was selected. In order to obtain relatively reliable analysis results, the study examined the time data of the hourly average temperature, hourly maximum temperature, and hourly minimum temperature in the world time and meteorological day boundary (20:00 Beijing time is the day boundary), and the comparative analysis showed that the amount of suspicious information and information of the two observation time systems were basically consistent.

[0105] Further manual verification shows that the erroneous information at two times at the beginning and end of a day (such as 00:00 and 23:00 UTC) is basically credible. Therefore, based on the UTC check results, for the highest and lowest temperatures, the erroneous hourly average temperatures that appeared at other times than 00:00 and 23:00 were set as missing (except for the hourly minimum temperatures at 00:00 on January 20, 2013 at A2169 station, 00:00 on November 5, 2009 at A2567 station, and 00:00 on March 28, 2011 at A3465 station). Finally, the missing hourly maximum temperature and hourly minimum temperature values ​​were 21 at 18 regional automatic weather stations and 66 at 31 regional automatic weather stations, respectively. As for the hourly average temperature, in addition to considering that the erroneous information at 00:00 and 23:00 is basically credible, the inspection results of the meteorological daily community were compared and it was found that there were some abnormal data at the A3364 regional automatic weather station during January 14-19, 2011, which were basically above 35.0℃, and even reached 50.0℃. However, due to the lack of measurement of the hourly maximum temperature or the hourly minimum temperature at the same time, no reasonable erroneous information was obtained in the internal consistency check. Therefore, the abnormal hourly average temperature values ​​during this period were all set as missing. Finally, there were 19 missing hourly average temperature values ​​for the three regional automatic weather stations.

[0106] Spatial consistency check: using formula (2), the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the inspected area is compared with the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the adjacent area for analysis. If it is confirmed to be erroneous data after manual verification, the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the inspected area is set as missing.

[0107] |Z t -Z t0 |>nσ t0 , (2)

[0108] Among them, Z t and Z t0 are the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the standardized series of the automatic weather station in the inspected area and the automatic weather station in the adjacent area, respectively; t0 It is the daily standard deviation of the standardized series of the automatic weather stations in the neighboring regions of the regional automatic weather station constructed by arithmetic averaging. The statistical principle of the daily series standard deviation is consistent with the above-mentioned climate anomaly inspection step.

[0109] The standard deviation multiple n = 3-5, the principle of selecting automatic weather stations in the neighboring area is: 3, 5, and 20 stations that are spatially close to the inspected station, and constructed by arithmetic averaging.

[0110] The data of 325 regional automatic weather stations were compared under the condition of no missing measurements, and the inspection results of hourly average temperature, hourly maximum temperature and hourly minimum temperature at 3, 4 and 5 times standard deviation were obtained. The error information and information of 5 times standard deviation were relatively reasonable; then, the inspection results of selecting 3, 5 and 20 neighboring regional automatic weather stations based on spatial distance were compared and analyzed under 5 times standard deviation, and the error information of fitting 3 stations basically included that of 5 and 20 regional automatic weather stations. Therefore, the standard deviation multiple n=5 was finally selected, and the number of neighboring regional automatic weather stations was 3.

[0111] Through further manual verification, the hourly average temperature, hourly maximum temperature and hourly minimum temperature data in the 5 times standard deviation, 3 stations to build the adjacent station inspection results, there are also two times (such as 00:00 and 23:00 UTC) at the beginning and end of the day. The erroneous information is basically credible. Therefore, in the end, all the erroneous information at 00:00 and 23:00 (UTC) in the hourly average temperature, hourly maximum temperature and hourly minimum temperature inspection results were removed, as well as |Z t -Z t0 |<5.0, and the remaining suspected erroneous information that was manually verified to be erroneous was set as missing. The suspected erroneous information of hourly average temperature, hourly maximum temperature and hourly minimum temperature set as missing were 5 for 3 regional automatic weather stations, 2 for 2 regional automatic weather stations, and 1 for 1 regional automatic weather station, as shown in Table 2.

[0112] Table 2 Correction results of spatial consistency error information

[0113]

[0114] Step 3: Effectively screen regional automatic weather stations based on data sources that have undergone quality control according to data integrity and time series continuity.

[0115] The step 3 comprises the following sub-steps:

[0116] Step 3.1: Select the regional automatic weather stations whose observation series of hourly average temperature, maximum temperature and minimum temperature are at least 10 years long and have continuous and valid observation data during this period.

[0117] Step 3.2: Among the regional automatic weather stations selected in step 3.1, further select regional automatic weather stations whose hourly average temperature, maximum temperature and minimum temperature meet the screening conditions of step 3.1.

[0118] Referring to the principle of statistics on the average value of 24 scheduled records in the Specifications for Ground Meteorological Observations (China Meteorological Administration, 2003), that is, if the 24 hourly records are missing 6 or more times in a day, no daily statistics will be made; if the records are missing 7 or more times in a month, no monthly statistics will be made; if the records are missing 1 or more times in a year, no annual statistics will be made. The number of effective regional automatic meteorological stations for the three types of temperature elements after quality control from 2008 to 2023 was counted year by year.

[0119] There was no relatively complete regional automatic weather station observation data in 2008 and 2011. Therefore, in order to ensure the continuity and integrity of the research data, in this embodiment, the 325 regional automatic weather stations after quality control were further screened based on the principle that the observation sequence length must be at least 10 years or more, and there must be continuous observation data from 2012 to 2023. 153 regional automatic weather stations with qualified hourly average temperatures and 114 regional automatic weather stations with qualified hourly maximum temperatures and hourly minimum temperatures were obtained. Furthermore, 112 regional automatic weather stations with qualified hourly average temperatures, hourly maximum temperatures, and hourly minimum temperatures were screened out as the basis for the next step of homogenization analysis.

[0120] Step 4: Perform homogenization analysis on the hourly average temperature, maximum temperature and minimum temperature of the screened regional automatic weather stations to obtain the homogenized hourly temperature data of the regional automatic weather stations.

[0121] The observation data of 112 regional automatic weather stations were converted from universal time to meteorological day boundaries (20:00 Beijing time is the day boundary), and the observation data of regional automatic weather stations from January 1, 2008 to December 31, 2023 were homogenized. The test results show that among the 112 regional automatic weather stations, the stations with no exact regional automatic weather station metadata support for the daily average temperature, daily maximum and daily minimum temperatures were detected to account for 13.4%, 37.5% and 33.0% of the stations, respectively. The statistical series heterogeneity breakpoints mainly appeared in 2013. Since the non-reference sequence method (PMFT) was used in the study for homogeneity test, this common series heterogeneity breakpoint largely reflects the abnormal changes in regional climate. Therefore, the original values ​​of these regional automatic weather stations were kept without processing.

[0122] The step 4 comprises the following sub-steps:

[0123] Step 4.1: Count the hourly average temperature, maximum temperature, and minimum temperature data into daily average temperature, daily maximum temperature, and daily minimum temperature, respectively. Use the Penalized Maximum F Test (PMFT) to test the homogeneity of the daily average temperature, daily maximum temperature, and daily minimum temperature series, with a significance level of α of 0.01, to obtain the breakpoints of the series heterogeneity.

[0124] Step 4.2: Verify the authenticity of the sequence heterogeneity breakpoints. Based on the metadata of the regional automatic weather stations, eliminate the statistically significant sequence heterogeneity breakpoints that are not supported by the exact regional automatic weather station metadata, retain the statistically significant sequence heterogeneity breakpoints that are supported by the exact regional automatic weather station metadata, and determine the regional automatic weather stations with such characteristics as heterogeneous regional automatic weather stations.

[0125] Step 4.3: Eliminate regional automatic weather stations with non-uniform characteristics to form the homogenized hourly temperature data of regional automatic weather stations.

[0126] Table 3-5 gives the information of regional automatic weather stations with serial inhomogeneity breakpoints for the hourly average temperature, hourly maximum temperature, and hourly minimum temperature obtained from the test and with definite regional automatic weather metadata support, as well as the statistical significance of their serial inhomogeneity breakpoints.

[0127] Table 3 Information of regional automatic weather stations with non-uniform hourly average temperature

[0128]

[0129]

[0130] Table 4. Regional automatic weather station information for non-uniform hourly maximum temperatures

[0131]

[0132] Table 5 Information of regional automatic weather stations with non-uniform hourly minimum temperatures

[0133]

[0134]

[0135] As can be seen from Table 3-5, station relocation is the main reason for the non-uniformity of the temperature observation series of regional automatic weather stations. Among them, the hourly minimum temperature series is most significantly affected by it, followed by the hourly average temperature series, while the hourly maximum temperature series is relatively least affected by it. This feature is also reflected in the homogeneity test of the temperature series of national meteorological stations.

[0136] From the test statistics of the heterogeneity breakpoints of each sequence, except for the hourly minimum temperature of the A2263 regional automatic weather station (Baodi District), the hourly average temperature, hourly maximum temperature and hourly minimum temperature PF max The values ​​all significantly exceed the confidence interval of α = 0.01. Therefore, although the hourly minimum temperature sequence of the A2263 regional automatic weather station is supported by the regional automatic weather station metadata and statistically significant at the time point of March 26, 2020, in order to reconfirm the rationality and reliability of the test results, the hourly minimum temperature sequence of the A2263 regional automatic weather station from 2008 to 2023 was manually verified in this embodiment. Through the analysis of the hourly data curve chart, it was found that the hourly minimum temperature value of -46.5°C at 06:00 on December 7, 2011 was erroneous data (but it was not checked out during the quality control process. The main reason was that there were only 3 observations on that day, and the others were missing). Therefore, the value was changed to missing, and the homogeneity test of the A2263 regional automatic weather station was re-performed. It was found that the sequence heterogeneity breakpoint on March 26, 2020 was not statistically significant (α = 0.01), so the regional automatic weather station was retained without processing.

[0137] Finally, in order to ensure the consistency of temperature data, the six regional automatic weather stations A2111 (Jizhou District), A2505 (Jinghai District), A2702 (Xiqing District), A2706 (Xiqing District), A3102 (Binhai New Area), and A3105 (Binhai New Area) with hourly average temperature, hourly maximum temperature, and hourly minimum temperature were deleted (that is, among the 112 regional automatic weather stations, the proportion of non-uniform regional automatic weather stations was 5.4%), forming a relatively uniform sequence of hourly average temperature, hourly maximum temperature, and hourly minimum temperature of regional automatic weather stations in Tianjin from 2008 to 2023 (a total of 106 regional automatic weather station files).

[0138] Step 5: Conduct data quality assessment on the homogenized hourly temperature data of the regional automatic weather station.

[0139] The step 5 comprises the following sub-steps:

[0140] Step 5.1: Count the hourly average temperature, maximum temperature and minimum temperature after the homogenization analysis in step 4 as daily values.

[0141] Step 5.2: Perform data quality assessment on the average sequence of the regional automatic weather station corresponding to the hourly average temperature, maximum temperature and minimum temperature of the daily value. The assessment index is the standard deviation of the difference sequence of the observation data between the regional automatic weather station and the national automatic weather station.

[0142] Step 5.3: Repeat steps 5.1 and 5.2 to evaluate the data quality of the automatic weather stations in each region after screening.

[0143] For ease of comparison, in this embodiment, the hourly temperature series of 106 regional automatic weather stations are converted from universal time to meteorological day boundary (20:00 Beijing time is the day boundary), and are counted as daily average temperature, daily maximum temperature and daily minimum temperature respectively. The complete time period of 2012-2023 in which the series is not interrupted is selected for analysis. At the same time, considering that the number of regional automatic weather stations in each administrative area is limited and the temperature elements have good climate continuity, the arithmetic mean is used to construct the average series of regional automatic weather stations in each administrative area. The observation data of national automatic weather stations in 11 administrative areas, including Jizhou (54428), Baodi (54525), Ninghe (54529), Wuqing (54523), Jinghai (54619), Xiqing (54527), Dongli (54526), ​​Jinnan (54622), Beichen (54528), Binhai New Area (54623) and Urban Area (54517) are used as reference values. The average series of regional automatic weather stations and the data quality of each regional automatic weather station are evaluated respectively, and the evaluation indicator is the standard deviation of the difference series of daily values ​​between regional automatic weather stations and national automatic weather stations.

[0144] The probability distribution of the standard deviation of the daily difference between the average series of regional automatic weather stations in the 11 administrative regions and the daily average temperature, daily maximum temperature, and daily minimum temperature of the corresponding national automatic weather stations: the difference in the standard deviation of the daily average temperature is relatively the smallest, followed by the daily maximum temperature, and the daily minimum temperature is relatively the largest. The standard deviation of the daily average temperature difference in the 11 administrative regions is within the range of 0.0-0.8℃, reaching more than 97.3% (except for the 54428 district), and the standard deviation of the daily maximum temperature and daily minimum temperature difference is within the range of 0.0-1.5℃, reaching more than 99.5% and 94.0% (except for the 54428 district). Among them, the daily average temperature and daily minimum temperature in the 54428 district are within the range of 0.0-1.5℃ and 0.0-2.0℃, reaching 100% and 93.2% respectively.

[0145] For the daily average temperature, the standard deviation ratio of the difference with the corresponding national automatic weather station is within the range of 0.0-1.5℃. Except for the four regional automatic weather stations A2165 (51.9%), A2113 (86.1%), A2109 (89.1%) and A2112 (89.6%) in Jizhou District (54428), the others all reach more than 95%, and the regional automatic weather stations reaching 100% account for about 80%. The metadata information of the regional automatic weather stations shows that the altitudes of these four stations are 555m, 330m, 230m and 140mm, respectively, which are significantly higher than the altitude of the corresponding national automatic weather stations in the district (32m). The standard deviation of the difference between the daily maximum temperature of each regional automatic weather station and the corresponding national automatic weather station is within the range of 0.0-2.0℃. Except for A2804 (82.8%) and A2661 (93.4%) in Binhai New Area (54623) and A2165 (92.3%) in Jizhou District (54428), the others all reach more than 95%, and the regional automatic weather stations reaching more than 99% account for 83%. The standard deviation of the difference of the daily minimum temperature is relatively the largest, and the A2165 regional automatic weather station in Jizhou District (54428) is the lowest at 63.1%. In addition, there are 5 regional automatic weather stations in the area with a proportion of 86.6%-92.3%, including A2112 (88.8%), A2109 (91.0%) and A2113 (92.1%). Similarly, the altitudes of these regional automatic weather stations are basically around 100m-230m. This also explains why the standard deviation of the difference between the daily average temperature and the daily minimum temperature in the 54428 district obtained in the above analysis is larger than that in other administrative districts. In addition, the standard deviation of the daily minimum temperature difference of the automatic weather stations in other regions is more than 95%, of which 83% are more than 99%.

[0146] Therefore, on the whole, the hourly average temperature, hourly maximum temperature and hourly minimum temperature data of 106 regional automatic weather stations in Tianjin obtained through data quality control and homogenization analysis have certain reliability.

[0147] Step 6: Analyze temperature trend characteristics.

[0148] Preferably, robust regression can be used to fit and analyze the trend variation amplitude and characteristics of the annual and seasonal series of the hourly average temperature, maximum temperature and minimum temperature of the regional automatic weather station.

[0149] This example uses robust regression (iterative weighted least squares method) (Cross, 1977; Shi Neng et al., 1992) to fit the trend changes of the annual and seasonal series of hourly average temperature, hourly maximum temperature and hourly minimum temperature of 106 regional automatic weather stations in Tianjin from 2021 to 2023. Robust regression analysis can ensure the correctness of regression parameter solution as much as possible under the premise that there may be gross errors (outliers, abnormal values, missing values ​​or multicollinearity, etc.) in the dependent variable samples, and can give relatively unbiased and effective parameter estimates compared with least squares estimation.

[0150] For the annual series change, 67.0% of the 106 regional automatic weather stations showed a significant increase trend (α = 0.05) in the hourly average temperature, with a range of 0.863-2.984℃ / 10yr, of which 94.4% reached 1.0℃ / 10yr or more. The hourly maximum temperature showed a significant increase trend (α = 0.05) in 79.2% of the regional automatic weather stations, with a range of 1.100-3.600℃ / 10yr, and 71.4% reached 1.5℃ / 10yr or more. There are 40.6% of regional automatic weather stations with a significant increasing trend (α=0.05) of the hourly minimum temperature, with a range of 0.660-2.606℃ / 10yr, of which 69.8% of the regional automatic weather stations reach above 1.0℃ / 10yr; in addition, 15.1% of the regional automatic weather stations show a decreasing trend, but from a statistically significant perspective, only the A2162 in Jizhou District and the A2202 in Baodi District have a statistically significant decreasing trend (α=0.05), with amplitudes of 1.467℃ / 10yr and 1.492℃ / 10yr, respectively.

[0151] From the perspective of seasonal series changes, hourly average temperature, 97.2% of regional automatic weather stations showed an increasing trend in spring, but only the A3506, A3510, and A3512 regional automatic weather stations in the urban area were statistically significant (α = 0.05), with amplitudes of 2.174℃ / 10yr, 1.330℃ / 10yr, and 1.428℃ / 10yr, respectively. In summer, 98.1% of regional automatic weather stations showed an increasing trend, of which 46.2% were statistically significant (α = 0.05), with an amplitude range of 0.939-3.014℃ / 10yr. Similarly, in autumn, 98.1% of regional automatic weather stations showed an increasing trend, of which 70.2% were statistically significant, with an amplitude range of 1.057-2.905℃ / 10yr. While 52.8% of the regional automatic weather stations showed an increasing trend in winter, only A3512 in the urban area was statistically significant (α=0.05), with a magnitude of 1.722℃ / 10yr; in addition, 46.2% of the regional automatic weather stations showed a decreasing trend, but only A2211 in Baodi District was statistically significant (α=0.05), with a magnitude of 1.945℃ / 10yr.

[0152] For the hourly maximum temperature, 92.5% of the regional automatic weather stations showed an increasing trend in spring, of which A2263 in Baodi District and A3512 in the urban area were statistically significant (α = 0.05), with amplitudes of 2.346℃ / 10yr and 2.308℃ / 10yr, respectively. In summer, 96.2% of the regional automatic weather stations showed an increasing trend, of which 39.2% were statistically significant (α = 0.05), with amplitudes ranging from 1.319 to 3.551℃ / 10yr; at the same time, A2309 in Ninghe District and A3302 in Jinnan District showed a significant decreasing trend (α = 0.05), with amplitudes of 1.710℃ / 10yr and 1.122℃ / 10yr, respectively. In autumn, 100% of the regional automatic weather stations showed an increasing trend, of which 63.2% were statistically significant (α = 0.05), with amplitudes ranging from 1.316 to 3.760℃ / 10yr. In winter, 84.0% of the regional automatic weather stations showed an increasing trend, but only 7.9% were statistically significant (α=0.05), with a range of 2.308-3.192℃ / 10yr.

[0153] For the hourly minimum temperature, 84.9% of the regional automatic weather stations showed an increasing trend in spring, and only 1.3% were statistically significant (α = 0.05), with a range of 0.964-2.493℃ / 10yr; in addition, A2118 and A2162 in Jizhou District showed a significant decreasing trend (α = 0.05), with amplitudes of 1.419℃ / 10yr and 0.713℃ / 10yr, respectively. In summer, 95.3% of the regional automatic weather stations showed an increasing trend, of which 33.7% were statistically significant (α = 0.05), with an amplitude range of 0.958-2.512℃ / 10yr. In autumn, 92.5% of the regional automatic weather stations showed an increasing trend, of which 45.9% showed statistically significant trend changes (α = 0.05), with a range of 1.030-2.581℃ / 10yr; while A2162 in Jizhou District showed a significant decreasing trend (α = 0.05), with a range of 2.583℃ / 10yr. In winter, 67.9% of the regional automatic weather stations showed a decreasing trend, of which A2263 in Baodi District, A2604 in Binhai New Area, and A3517 in the urban area were statistically significant (α = 0.05), with ranges of 2.017℃ / 10yr, 2.889℃ / 10yr, and 1.889℃ / 10yr, respectively.

[0154] Therefore, based on the analysis of the observation data of the automatic weather stations in the above regions, from the perspective of annual scale changes, the increase trend and change range of the hourly maximum temperature in Tianjin are relatively obvious, followed by the hourly average temperature, while the hourly minimum temperature is relatively the smallest. For the seasonal scale, the change characteristics of the hourly average temperature, hourly maximum temperature, and hourly minimum temperature are basically consistent with the annual scale. Among them, the temperature increase trend and change range in summer and autumn are relatively significant, especially in autumn.

[0155] This example is based on the hourly average temperature, hourly maximum temperature and hourly minimum temperature observation data of 325 regional automatic weather stations in Tianjin, combined with the regional automatic weather station metadata information such as station relocation, element upgrade and collector update since the establishment of the station, through data quality control and homogenization analysis, to form a relatively uniform hourly temperature dataset of regional automatic weather stations in Tianjin from 2008 to 2023. The main conclusions are as follows:

[0156] (1) The data quality control results show that the hourly average temperature, hourly maximum temperature and hourly minimum temperature mainly have problems with internal consistency, climate anomalies and spatial consistency. By consulting the metadata of regional automatic weather stations and analyzing the change characteristics of temperature time series, 148 suspected internal consistency errors of 15 regional automatic weather stations were omitted or corrected through manual verification; for the problem of climate anomalies, under the condition of a threshold of 4 times the standard deviation, 19 suspected errors of hourly average temperature, hourly maximum temperature and hourly minimum temperature were omitted for 3 regional automatic weather stations, 21 for 18 regional automatic weather stations and 66 for 31 regional automatic weather stations; for the results of spatial consistency check, under the conditions of a threshold of 5 times the standard deviation and 3 adjacent regional automatic weather stations, 5 for hourly average temperature, hourly maximum temperature and hourly minimum temperature were omitted for 3 regional automatic weather stations, 2 for 2 regional automatic weather stations and 1 for 1 regional automatic weather station.

[0157] (2) Based on the principle that the length of the observation time series should be at least 10 years or more and there should be continuous observation data from 2012 to 2023, 112 regional automatic weather stations with common hourly average temperature, hourly maximum temperature and hourly minimum temperature after quality control were selected. On this basis, the penalized maximum F test (PMFT) was used to test the homogeneity of the daily data of the 112 regional automatic weather stations in combination with the metadata of the regional automatic weather stations. It was found that station migration was the main reason for the inhomogeneity of the temperature observation series of the regional automatic weather stations. Among them, the minimum temperature series was most significantly affected, followed by the average temperature series, while the maximum temperature series was relatively least affected.

[0158] (3) Then, six regional automatic weather stations with statistically significant breakpoints in hourly average temperature, hourly maximum temperature and hourly minimum temperature and definite regional automatic weather station metadata support were eliminated, and finally relatively uniform hourly average temperature, hourly maximum temperature and hourly minimum temperature series of 106 regional automatic weather stations in Tianjin from 2008 to 2023 were obtained. By comparing and analyzing the regional average series from 2012 to 2023 and the standard deviation of the daily value difference between each regional automatic weather station and the corresponding 11 administrative regions, it was found that the hourly temperature data of the 106 regional automatic weather stations developed were relatively reliable.

[0159] (4) Analysis of observation data from regional automatic weather stations shows that in the past 10 years, the increasing trend and variation range of the hourly maximum temperature in Tianjin are relatively obvious, followed by the hourly average temperature, while the hourly minimum temperature is relatively the smallest. In addition, the increasing trend and variation range of temperature in summer and autumn are the most significant, especially in autumn.

[0160] It should be pointed out that in the homogenization analysis of the observation data of regional automatic weather stations, the most difficult part is the collection and organization of metadata of regional automatic weather stations. Without relatively detailed metadata of regional automatic weather stations, it will largely make it impossible to objectively judge the reliability of some statistically significant breakpoints. It may also make it impossible to find regional automatic weather stations with relatively good uniformity as objective basis for the reference sequence of the stations to be tested. Therefore, in this embodiment, in the presence of reasonable regional automatic weather station metadata, only non-uniform regional automatic weather stations are eliminated without correction, and in the absence of a reference sequence, the test statistic PF of the statistically significant breakpoints is used to eliminate the non-uniform regional automatic weather stations. max The uniformity of regional automatic weather stations can be further determined by analyzing the values. Therefore, in the future development of homogenized datasets of regional automatic weather station observation sequences, the collection and collation of detailed metadata of regional automatic weather stations is still a top priority, so as to achieve the uniformity correction of long-term climate observation data by establishing a reasonable and reliable reference sequence, and provide a more complete and reliable data foundation for the research of regional / local climate and extreme climate change.

[0161] The above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing homogenized hourly temperature data of a regional automatic weather station, characterized by: The following steps are involved: Step 1: Collect data sources of automatic weather stations in each region and observation data of national automatic weather stations; The data sources of regional automatic weather stations include regional automatic weather station observation data and regional automatic weather station metadata; Step 2: Perform data quality control on the regional automatic weather station observation data collected in step 1, and use climate limit value check, internal consistency check, negative value check of June-August temperature, climate anomaly check and spatial consistency check to control the quality of the regional automatic weather station observation data; Step 3: Effectively screen regional automatic weather stations based on data sources that have undergone quality control according to data integrity and time series continuity; Step 4: Perform homogenization analysis on the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the selected regional automatic weather stations to obtain the homogenized hourly temperature data of the regional automatic weather stations; Step 5: Conduct data quality assessment on the homogenized hourly temperature data of the regional automatic weather station; Step 6: Analyze temperature trend characteristics.

2. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: The regional automatic weather station is an unmanned station; the observation data of the regional automatic weather station includes the hourly average temperature, hourly maximum temperature and hourly minimum temperature of the regional automatic weather station since its establishment; The metadata of the regional automatic weather station includes the time of relocation, element upgrade time, and collector update time of the regional automatic weather station since its establishment; The national automatic weather station is a manned station, and the observation data of the national automatic weather station include the daily average temperature, daily maximum temperature and daily minimum temperature of the national automatic weather station after uniformity correction since its establishment.

3. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: In the step 2, the specific operation of the climate limit value check is: set the hourly average temperature outside the range of -80℃-60℃ as missing, set the hourly maximum temperature outside the range of -80℃-60℃ as missing, and set the hourly minimum temperature outside the range of -80℃-60℃ as missing.

4. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: In the step 2, the specific operation of the internal consistency check is: set the judgment logic: hourly minimum temperature ≤ hourly average temperature ≤ hourly maximum temperature. For the regional automatic weather station observation data that does not meet the judgment logic, if the regional automatic weather station observation data is reasonable in its own time series changes, the hourly maximum temperature or the hourly minimum temperature is modified based on the hourly average temperature. However, if it is manually verified to be erroneous data, it is set to missing.

5. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: In the step 2, the specific operation of checking whether the temperature values ​​from June to August are negative is: when the observation data of the regional automatic weather station from June to August are negative, the observation data of the regional automatic weather station are set as missing.

6. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: In step 2, the specific operation of checking the climate anomaly is: using formula (1), comparing the observation data of the regional automatic weather station with the corresponding daily standard deviation temperature value, and when it is manually verified to be erroneous data, setting the observation data of the regional automatic weather station as missing; Among them, i is a day, j is an hour on a day, and x is ij The hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked. is the average value of the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the automatic weather station in the area being checked, σ i is the standard deviation of the daily series of hourly average temperature, hourly maximum temperature or hourly minimum temperature of the regional automatic weather station being checked, n is the multiple of standard deviation, n = 3-5; If the hourly average temperature, hourly maximum temperature or hourly minimum temperature of the day is missing for 6 times or more, the statistical value of the daily series mean or standard deviation will be regarded as missing and no statistical check will be performed.

7. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: In step 2, the specific operation of the spatial consistency check is: using formula (2), the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is compared and analyzed with the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the adjacent area respectively. If it is manually verified to be erroneous data, the hourly average temperature or hourly maximum temperature or hourly minimum temperature of the automatic weather station in the inspected area is set as missing; |With t -Z t0 |>nσ t0 , (2) Among them, Z t and Z t0 are the hourly average temperature, hourly maximum temperature, or hourly minimum temperature of the standardized series of the automatic weather station in the inspected area and the automatic weather station in the adjacent area, respectively; t0 The standard deviation of daily values ​​of the normalized series of the neighboring regional automatic weather stations constructed by arithmetic averaging.

8. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: The step 3 comprises the following sub-steps: Step 3.1: Select the regional automatic weather stations whose observation series of hourly average temperature, hourly maximum temperature and hourly minimum temperature are at least 10 years long and have continuous and valid observation data during this period; Step 3.2: Among the regional automatic weather stations selected in step 3.1, further select regional automatic weather stations whose hourly average temperature, hourly maximum temperature and hourly minimum temperature meet the selection conditions of step 3.

1.

9. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: The step 4 comprises the following sub-steps: Step 4.1: Count the hourly average temperature, hourly maximum temperature, and hourly minimum temperature data into daily average temperature, daily maximum temperature, and daily minimum temperature, respectively. Use the maximum penalty F test to test the homogeneity of the daily average temperature, daily maximum temperature, and daily minimum temperature series, with a significance level of α of 0.01, to obtain the breakpoints of the series heterogeneity. Step 4.2: Verify the authenticity of the sequence heterogeneity breakpoints. Based on the metadata of the regional automatic weather station, remove the statistically significant sequence heterogeneity breakpoints that are not supported by the exact regional automatic weather station metadata, retain the statistically significant sequence heterogeneity breakpoints that are supported by the exact regional automatic weather station metadata, and determine the regional automatic weather stations with such characteristics as heterogeneous regional automatic weather stations; Step 4.3: Eliminate regional automatic weather stations with non-uniform characteristics to form the homogenized hourly temperature data of regional automatic weather stations.

10. The method for processing homogenized hourly temperature data of a regional automatic weather station according to claim 1, characterized in that: The step 5 comprises the following sub-steps: Step 5.1: Count the hourly average temperature, hourly maximum temperature and hourly minimum temperature after homogenization analysis in step 4 as daily values; Step 5.2: Perform data quality assessment on the average sequence of the regional automatic weather station corresponding to the daily hourly average temperature, hourly maximum temperature and hourly minimum temperature. The assessment index is the standard deviation of the difference sequence of observation data between the regional automatic weather station and the national automatic weather station. Step 5.3: Repeat steps 5.1 and 5.2 to evaluate the data quality of the automatic weather stations in each region after screening.

Images