Dataset Development on Centennial Homogenized Monthly/Yearly Mean Temperature at Jiujiang Meteorological Station, Jiangxi Province of China (1924?C2023)

Zhan, L. F.^1,2  Dong, B. H.^1,2  Xu, B.^1,2*   Li, Y.^1,2   Xin, J. J.^1,2   Wang, L. Y.³

1. Jiangxi Provincial Climate Center, Nanchang 330096, China;

2. Nanchang National Climate Observatory, Nanchang 330200, China;

3. Jiangxi Vocational and Technical College of Information Application, Nanchang 330043, China

Abstract: Long-term homogenized observational time series are crucial for accurate assessment and attribution of climate change. However, at various times, most meteorological stations in China are affected by factors such as station relocation, instrument replacement, and environmental change, resulting in heterogeneity in observational data series. In this study, based on multisource monthly mean temperature data, the monthly mean temperature data recorded at Jiujiang Meteorological Station in Jiangxi Province (China) during 1924?C2023 were interpolated using the standardized sequential method, taking data from Wuhan Meteorological Station as reference. The interpolated data were subjected to a homogenization test using the penalized maximal F test method, and then corrected using the quantile matching method. Thus, the Centennial homogenized monthly/yearly mean temperature dataset of Jiujiang Meteorological Station, Jiangxi Province, China (1924?C2023) was constructed. Comparative analysis with centennial homogenized temperature data of neighboring meteorological stations revealed a correlation coefficient of >0.9, verifying that the construction method is scientific and that this dataset has certain reliability. The dataset is archived in .txt data format, and consists of 2 data files with data size of 16.5 KB.

Keywords: Jiujiang; temperature; centennial series; interpolated; homogenized

DOI: https://doi.org/10.3974/geodp.2025.01.10

Dataset Availability Statement:

The dataset supporting this paper was published and is accessible through the Digital Journal of Global Change Repository at: https://doi.org/10.3974/geodb.2024.12.03.V1.

1 Introduction

In the context of global climate change, study of regional climatic variability is of great importance for improved understanding of the spatial characteristics of climate change and the associated potential impacts^[1,2]. The East Asian monsoon region is one area that is highly sensitive to the effects of global climate change, where regional climate change is driven directly by global warming and multiple other factors such as the regional atmospheric circulation and changes in land use cover^[3]. Jiujiang is an important city in the middle-lower reaches of the Yangtze River (China) that has typical East Asian monsoon climate characteristics. Its long-term climate records provide a valuable basis for studying the characteristics of the regional climate and their relationship with global climate change. However, owing to the diverse sources of historical meteorological observational data, a number of missing measurements, and the inhomogeneity inherent in the time series, high-quality datasets that accurately reflect the long-term trend of climate change in the Jiujiang region are scarce^[4].

Constructing continuous, reliable, and long-term time series of homogenized meteorological data forms the basis of climate change research^[5,6]. Homogenization is the process of systematically correcting and unifying multisource meteorological observational data to eliminate systematic differences and nonhomogeneous characteristics between different data elements so that they reflect the true characteristics of climate change. In recent years, many studies have achieved remarkable results using homogenization techniques to correct historical meteorological data^[7?C10]. However, in specific applications, it remains necessary to combine regional characteristics and historical data characteristics for more detailed and targeted processing.

In this study, taking the Jiujiang Meteorological Station (Jiangxi Province, China) as the research object, multisource meteorological data from 1924?C1938 and from 1951?C2023 were integrated with homogenized data from neighboring stations from 1924?C2016. Then, a homogenized monthly temperature dataset for Jiujiang Meteorological Station from 1924?C2023 was constructed using the standardized sequential method, a homogenization test, and breakpoint correction technology. Finally, the temperature change characteristics at Jiujiang Meteorological Station over the past 100 years were revealed. This study provides important data support for climate change research in the Jiujiang region, and represents a reference for the processes of homogenization and sequence construction of meteorological data in other regions.

2 Metadata of the Dataset

The name, authors, geographical region, data period, temporal resolution, spatial resolution, dataset composition, data publication and sharing service platform, and data sharing policies of the Centennial homogenized monthly/yearly mean temperature dataset of Jiujiang Meteorological Station, Jiangxi Province, China (1924?C2023)^[11] are provided in Table 1.

3 Data Development Methods

3.1 Data Sources

The Jiujiang Meteorological Station data from January 1924 to March 1938 were obtained from the ??China Temperature Data??^[13] record compiled by the Joint Data Office of the

Table 1  Metadata summary of the Centennial homogenized monthly/yearly mean temperature dataset at Jiujiang Meteorological Station, Jiangxi Province, China (1924?C2023)

China Central Meteorological Administration and the Institute of Geology and Geophysics (Chinese Academy of Sciences), which was published in 1954. The data from 1951?C2023 were extracted from the ??Tianqing?? meteorological big data cloud platform^[14] developed by the National Meteorological Information Center. The centennial homogenized data of neighboring reference stations were taken from the dataset established by the Institute of Atmospheric Physics of the Chinese Academy of Sciences, after correcting the centennial homogenized monthly temperature series of 32 stations in China. This dataset well represents the characteristics of large-scale climate change in China over the past century^[15]. The geographical location of Jiujiang Meteorological Station is shown in Figure 1.

3.2 Algorithm Principles

3.2.1 Data Interpolation

The temperature series of Jiujiang Meteorological Station prior to 1951 was constructed

based on monthly mean temperature data in the ??China Temperature Data??. However, the temperature data record of Jiujiang Meteo­rological Station after preliminary integration showed continuous missing measurements from April 1938 to December 1950, with a rate of missing data up to 12.8%. To restore data integrity, the missing data were interpolated using the standardized sequential method, and neighboring stations with complete long-term time series sequences were taken as reference stations.

The calculation under the standardized sequential method can be expressed as follows^[16]:

                                                                                                   (1)

                                                                                                       (2)

                                                                                                  (3)

where denotes the standardized sequence,  denotes the mean standardized sequence of the reference stations, represents the data to be interpolated in the i^th month, j denotes the j^th reference station, refers to the data of the i^th month of the j^th station,  and are the multiyear mean and the standard deviation of the data of the i^th month of the j^th station, respectively, n represents the number of reference stations, and  and  are the multiyear mean and the standard deviation of the data of the i^th month of the station to be interpolated, respectively.

According to the requirements of reference^[17], the Jiujiang Meteorological Station should be used as the reference when selecting neighboring stations, and the data of neighboring meteorological stations within a horizontal distance of 300 km should be selected. Regarding the selected meteorological station, the starting observation year must be earlier than 1924, the data integrity must be high, and the site elevation should be similar to that of Jiujiang Meteorological Station. After comprehensive consideration, Wuhan Meteorological Station was finally selected as the reference station for the neighboring stations.

3.2.2 Homogenization Test and Correction

(1) Breakpoint test method

It was not until the early 1950s that China had relatively complete and systematic observational data. Prior to 1951, owing to the diverse sources of data recorded by China??s meteorological observation stations, the continuity, consistency, and standardization of the time series were inadequate. The lack of a reliable basis on which to evaluate the rationality of nonhomogeneous test results before 1951 further increased the difficulty of constructing a reference time series that truly reflected local climate change.

RHtest V4 software package is a statistical tool designed for homogeneity testing and adjustment of climate data. It is primarily used to detect and correct biases in climatic time series data (e.g., temperature, precipitation, etc.) caused by non-climatic factors (such as station relocations, instrument changes, observation method updates, etc.), thereby improving the reliability and consistency of the data. The penalized maximum F test in the RHtest V4 software package is a test method that can be performed without a reference sequence. Its primary advantage is that it can substantially reduce test bias caused by nonhomogeneous reference sequences and incomplete metadata information. Previous studies fully demonstrated the effectiveness and reliability of this method. Therefore, based on the monthly mean temperature series, this study adopted the penalized maximum F test method to conduct a homogenization test on the monthly mean temperature series of Jiujiang Meteorological Station from 1924?C2023. The test procedure followed the method discussed in reference^[18].

(2) Breakpoint correction method

The quantile matching method^[19] was used to correct the monthly data series. This method can ensure that the segments in the sequence to be tested match each other in the empirical distribution after eliminating the linear trend. In this study, 2 types of breakpoints were corrected: those before 1954 and those after 1954.

1) Breakpoints before 1954. Detected in both the yearly and the monthly series, these breakpoints represented notably discontinuous points. However, owing to lack of supporting detailed metadata, the correction positions were determined based on the breakpoint times detected in the monthly data series.

2) Breakpoints detected after 1954. Found in the yearly series or in the monthly series, these breakpoints were identified with supported from corresponding metadata. If the time of occurrence of these breakpoints differed from the information recorded in the station metadata by no more than one year, the position of the breakpoint was replaced and adjusted according to the specific time of the metadata record^[6].

The test adopted a 95% confidence level to ensure the reliability of the test results. This method is helpful for reducing test errors caused by nonhomogeneous reference sequences and lack of detailed metadata information, thereby improving the accuracy of the sequence correction.

3.3 Technical Roadmap

The technical roadmap for the development of the Centennial homogenized monthly/yearly mean temperature dataset of Jiujiang Meteorological Station, Jiangxi Province, China (1924?C2023) is shown in Figure 2. The meteorological observational data from 1924?C1938 and from 1951?C2023 were integrated with the statistically homogenized correction data from 1924?C2016. Then, the standardized sequential method was used to interpolate missing data in the Jiujiang Meteor­ological Station record. Based on the homogeni­zation test and the correction procedure, the yearly homogenized meteorological data series of the Jiujiang Meteorological Station from 1924?C2023 was constructed.

4 Data Results and Validation

4.1 Dataset Composition

This dataset contains monthly/yearly mean temperature data of the Jiujiang Meteorological Station from 1924?C2023. The monthly average temperature lists the data for year, month, and corresponding monthly average temperature (??C); the annual average temperature lists the year and corresponding annual average temperature (??C). The dataset is archived in the .txt format, and consists of 2 data files with data size of 16.5 KB.

4.2 Data Results

In this study, a complete centennial time series temperature dataset was constructed. As shown in Figure 3, prior to the 1950s, the yearly mean temperature at Jiujiang Meteorological Station showed a downward trend, whereas from the 1950s to the early 1990s there was no obvious change. From the 1990s, the yearly mean temperature began to show a clear upward trend. Overall, the rate of change during 1924?C2023 was approximately 1.0 ??/100a. As shown in Figure 4, obvious seasonal changes occurred in the distribution of the monthly mean temperature at Jiujiang Meteorological Station. The temperature was markedly higher in summer (June?CAugust) and notably lower in winter (December?CFebruary), showing typical monsoon climate characteristics. Additionally, from the late 20th century to the early 21st century, both the duration and the intensity of high temperatures in summer increased. This change is consistent with the characteristics of regional climate change under the background of global warming.

The trend of temperature change in different periods and the results of the statistical significance test are shown in Figure 5. Red columns indicate significant temperature change at the significance level of ??=0.05; blue columns indicate temperature change that did not reach the significance level. It is evident from Figure 5 that temperature showed a statistically significant upward trend during January?CJune (especially during February?CApril), October?CNovember, and throughout the entire year. This may be related to the intensified trend of climate warming, with faster warming in winter and spring, and delayed heat release in autumn, leading to significant temperature increases during these periods. During July?CSeptember and in December, the temperature changed only a little, and it even showed a downward trend in August. In these periods, the change trend did not reach the significance level. This is likely due to the fact that in summer, the region is under the control of stable weather systems such as the subtropical high, resulting in relatively small interannual variations in temperature. In December, which is early winter, cold air activities are frequent but their intensity and duration vary from year to year, leading to non-significant temperature changes.

4.3 Data Validation

The correlation coefficient, root mean square error, and standard deviation can reflect the performance of this dataset in different periods. As shown in Table 2, the correlation coefficient of this dataset for the annual average data from 1924 to 2023 was 0.94, indicating strong correlation between this dataset and the reference station. The full-year root mean square error of 0.27 indicates that the mean deviation between this dataset and the reference station was small; the standard deviation of 0.59 indicates that the volatility of the full-year data was low.

For the monthly data, the range of the correlation coefficient was 0.88?C0.98, and the correlation coefficient of the winter half year was higher than that of the summer half year, indicating that this dataset has stronger correlation with the reference station in the winter half year. The root mean square error range was 0.46?C1.13, with the smallest (largest) value in spring (winter), indicating that the value in spring is closest to that of the reference station, while the error is largest in winter. This is due to the fact that in spring, temperatures gradually warm up and the changes are relatively smooth, resulting in smaller root-mean-square errors (RMSE) compared to the reference station. In winter, temperatures are lower and are affected by cold air, leading to greater temperature fluctuations and larger RMSE compared to the reference station. The standard deviation range was 0.97?C1.9, with the lowest (highest) value in June (February), indicating less (more) fluctuation of the data.

Table 2  Comparison of monthly and yearly mean temperatures of Jiujiang Meteorological Station  (1924?C2023) with those of reference stations

5 Discussion and Conclusion

In this study, a homogenized monthly/yearly mean temperature dataset for 1924?C2023 was constructed for Jiujiang Meteorological Station in Jiangxi Province, China. The ??China Temperature Data?? record, ??Tianqing?? meteorological big data cloud platform of the National Meteorological Information Center, and centennial homogenized data of neighboring stations were comprehensively applied to interpolate missing data using the standardized sequential method. Additionally, RHtest V4 software was used to test and correct the homogeneity of the sequence, thereby ensuring continuity and reliability of the time series. Verification results showed that this dataset has strong correlation with the reference station (yearly correlation coefficient: 0.94) and a small root mean square error (0.27). Therefore, it can better reflect the characteristics of climate change in the Jiujiang region over the past century.

This study found that temperature change at the Jiujiang Meteorological Station has obvious seasonal characteristics, with statistically significant higher temperatures in summer (June?CAugust) and statistically significant lower temperatures in winter (December?C February), reflecting the typical characteristics of the East Asian monsoon climate. From the long-term trend (except in August), the mean temperature at Jiujiang Meteorological Station in each month showed an upward trend, especially during February?CApril when the most statistically significant temperature rise occurred, which is consistent with the regional characteristics of large temperature rise in spring and winter under the background of global warming. However, the temperature change during July?CSeptember was small, especially in August when there was a downward trend, and it failed to reach the significance level, which might reflect the complexity of the region being affected by local climate change or monsoon variability in summer. Additionally, both the duration and the intensity of high temperatures in summer from the late 20th century to the early 21st century increased, further indicating that the regional climate is being affected profoundly by global warming.

Despite major progress being achieved, this study had certain limitations. For example, owing to the diverse sources and insufficient metadata information of meteorological data prior to 1951, the verification and correction of breakpoints depended on the statistical characteristics of the series, which might have had certain impact on the accuracy of the results. Moreover, the potential impact of extreme climatic events on data homogeneity needs further exploration. In future studies, higher-resolution climate simulations should be combined with observational data to examine the driving mechanism and response characteristics of regional climate change, thereby providing more precise scientific support for the formulation of policies intended to address climate change.

Author Contributions

Xu, B. and Dong, B. H. completed the overall design for the development of the dataset. Zhan, L. F. collected and processed the Jiujiang temperature data, and designed the model and the algorithm. Zhan, L. F. and Li, Y. completed the data verification. Zhan, L. F., Xin, J. J. and Wang, L. Y. wrote the data paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]        Arnella, N. W., Lowe, J. A., Challinor, A. J., et al. Global and regional impacts of climate change at different levels of global temperature increase [J]. Climatic Change, 2019, 155(3): 377?C391.

[2]        Pan, X. D. A review of future climate change research based on regional climate models [J]. Plateau Meteorology, 2018, 37(5): 1440?C1448.

[3]        Ren, G., Chan, J. C. L., Kubota, H., et al. Historical and recent change in extreme climate over East Asia [J]. Climatic Change, 2021, 168(3): 1?C19.

[4]        Zhang, J. L., Liu, B., Ren, S. Q., et al. A 4 km daily gridded meteorological dataset for China from 2000 to 2020 [J]. Scientific Data, 2024, 11(1): 1230.

[5]        Tang, G. L., Ding, Y. H., Wang, S. W., et al. Comparative analysis of temperature curves in China over the past century [J]. Advances in Climate Change Research, 2009, 5(2): 71?C78.

[6]        Si, P., Hao, L. S., Luo, C. J., et al. Interpolation and homogeneity correction of missing temperature records at Baoding Meteorological Station in Hebei [J]. Advances in Climate Change Research, 2017, 13(1): 41?C51.

[7]        Li, Q. X. Current status and prospects of homogeneity research in Chinese climate data [J]. Meteorological Science and Technology, 2016, 6(3): 67?C74.

[8]        Shen, Y. L. Construction and evaluation of a high-resolution meteorological element grid dataset for Qinghai Province [J]. Meteorological Science and Technology, 2023, 13(4): 29?C33.

[9]        Wang, H. J., Tu, S. Y., Chen, Z. H. Experiment and error analysis of interpolation methods for missing daily temperature data [J]. Meteorology, 2008, 34(7): 83?C91.

[10]     Wang, Y. P. Comparison and homogeneity analysis of observation data from old and new meteorological stations in Gansu Province [J]. Plateau Meteorology, 2023, 42(2): 506?C514.

[11]     Zhan, L. F., Xu, B., Dong, B. H., et al. Centennial homogenized monthly/yearly mean temperature dataset at Jiujiang Meteorological Station, Jiangxi Province of China (1924?C2023) [J/DB/OL]. Digital Journal of Global Change Data Repository, 2024. https://doi.org/10.3974/geodb.2024.12.03.V1.

[12]     GCdataPR Editorial Office. GCdataPR data sharing policy [OL]. https://doi.org/10.3974/dp.policy.2014.05 (Updated 2017).

[13]     Central Weather Bureau. China Temperature Data [M]. Beijing: Central Weather Bureau, 1954.

[14]     China Meteorological Administration. CAAC meteo data services [OL]. http://www.weatherdt.com. cn/ datasource.html.

[15]     Institute of Atmospheric Physics, Chinese Academy of Sciences. CAAC meteo data services [OL]. http://www.weatherdt.com.cn/datasource.html.

[16]     Li, Z., Yan, Z. W., Cao, L. J., et al. Further-adjusted long-term temperature series in China based on MASH [J]. Advances in Atmospheric Sciences, 2018, 35(8): 909?C917.

[17]     Yu, Y., Li, J., Ren, Z. H., et al. Application of standard series method in interpolation of missing daily average temperature data [J]. Meteorology, 2012, 38(9): 1135?C1139.

[18]     Si, P., Xie, Y. Y. Homogeneity analysis of total solar radiation data in Tianjin [J]. Climate and Environment Research, 2015, 20(3): 269?C276.

[19]     Wang, X. L., Chen, H., Wu, Y., et al. New techniques for the detection and adjustment of shifts in daily precipitation data series [J]. Journal of Applied Meteorology and Climatology, 2010, 49(12): 2416?C2436.