Development of Gridded Dataset of Extreme Temperature Index in China Based on ETCCDI

Chen, Q. Y.^1,2   Zhang, Y.^1,2*  Liu, X. Y.^2,3  Lian, Q. L.^2,3  Xu, J. J.^1,4

1. South China Sea Institute of Marine Meteorology, Guangdong Ocean University, Zhanjiang 524088, China;

2. Laboratory for coastal Ocean Variation and Disaster Prediction, College of Ocean and Meteorology, Guangdong Ocean University, Zhanjiang 524088, China;

3. Key Laboratory of Climate, Resources and Environment in Continental Shelf Sea and Deep Sea, Guangdong Ocean University, Zhanjiang 524088, China;

4. Shenzhen Institute of Guangdong Ocean University, Shenzhen 518120, China

Abstract: Extreme weather and climatic events are occurring frequently in the context of global warming. To unify the definitions of extreme climate events in different countries and regions, the World Meteorological Organization (WMO) established the Expert Group on Climate Change Detection and Index (ETCCDI), which provides 26 representative extreme temperature and precipitation indices to regulate research on global extreme climate events. This study used the Dataset of daily values of basic meteorological elements of national surface meteorological stations in China (V3.0) to calculate 16 extreme temperature indices defined by ETCCDI in the Chinese region, including the hottest day (TXx), coldest day (TXn), warmest night (TNx), coldest night (TNn), warm days (TX90p), cool days (TX10p), warm nights (TN90p), cool nights (TN10p), summer days (SU), tropical nights (TR), ice days (ID), frost days (FD), warm spell duration index (WSDI), cold spell duration index (CSDI), diurnal temperature range (DTR), and growing season length (GSL). For the convenience of researchers in different disciplinary fields, this study adopted the angular distance weighting (ADW) method to interpolate station data into a spatial resolution of 0.25????0.25?? longitude and latitude grid, stored in .NetCDF format from 1961 to 2020 and with an annual or monthly temporal resolution. The total data amount was 1.43 GB, compressed to 353.0 MB. This dataset was named SimmEX_1961-2020_1.0. In addition to providing data support for the study of various extreme temperature event characteristics, it also has broad application prospects in the fields of the environment, economy, and energy.

Keywords: climate change; ETCCDI; extreme climate index; SimmEX

DOI: https://doi.org/10.3974/geodp.2024.01.08

CSTR: https://cstr.escience.org.cn/CSTR:20146.14.2024.01.08

Dataset Availability Statement:

The dataset supporting this paper was published and is accessible through the Digital Journal of Global Change Data Repository at: https://doi.org/10.3974/geodb.2024.05.06.V1 or https://cstr.escience.org.cn/CSTR:20146.11.2024.05.06.V1.

1 Introduction

The IPCC AR6^[1] shows that, in the context of global warming, the frequency and intensity of extreme weather and climate events, such as extreme heat events, heavy rainfall, and droughts, are increasing globally. This is having a serious impact on the development of human societies^[2] and the balance of ecosystems^[3]. Recognizing the characteristics of these extreme events and formulating scientific responses to them are of great significance for disaster prevention and reduction. An extreme climate index dataset with high spatial and temporal coverage and resolution was the basis for the above research. Although many scholars and institutions have produced extreme climate index datasets, such as Wang et al.^[4], who calculated the extreme temperature index in the coastal areas of China, and Ma et al.^[5] and Zhou et al.^[6], who produced extreme precipitation datasets for the Tibetan Plateau and Wei River Basin, respectively. These datasets have played an important role in local or regional research. However, to study the entire region of China, the spatial coverage of the dataset must be expanded further. The US National Aeronautics and Space Administration and UK Met Office??s Hadley Center launched their own global extreme climate index products, M2SMNXEDI and HadEX, respectively. Although they cover the entire Chinese region, these extreme index products were created using fewer original observation stations (approximately 400?C500 stations) and have lower spatial resolution (1.25????1.875??)^[7], which limits their reliability and applicability.

In order to obtain a high-spatiotemporal-resolution dataset of extreme temperature indices in China, we first utilized the Python programming language and the daily maximum (minimum) temperature and average temperature of 2,481 ground meteorological stations provided by the National Meteorological Information Center of the China Meteorological Administration to calculate 16 extreme temperature indices (Table 1) specified by the Expert Team on Climate Change Detection and Indices (ETCCDI). Then, a grid dataset of extreme temperature indices in China from 1961 to 2020 was generated using the angular distance

Table 1  16 extreme temperature indices

weighting (ADW) method. The dataset, named SimmEX_1961-2020_1.0 and abbreviated as SimmEX, was created by the Extreme Climate Research Group of the South China Sea Institute of Marine Meteorology of Guangdong Ocean University (GDOU-SIMM). The dataset provides fundamental data support for the study of extreme temperature events across China or at finer spatiotemporal scales, which will help to enhance people??s understanding of extreme temperature events. Meanwhile, it also has significant applications in climate modeling and scientific responses to climate change in the fields of the environment, economy, and energy.

2 Metadata of the Dataset

The metadata of the Grid dataset of extreme temperature index in China (1961?C2020) (V1.0)^[8] dataset is summarized in Table 2. It includes the dataset full name, short name, authors, year of the dataset, temporal resolution, spatial resolution, data format, data size, data files, data publisher, and data sharing policy, etc.

Table 2  Metadata summary of Grid dataset of extreme temperature index in China (1961?C2020) (V1.0)

3 Methods

3.1 Data Collection

The foundational meteorological data for the development of this dataset were obtained from the Dataset of daily values of basic mete­orological elements of national surface meteorological stations in China (V3.0), produced by the National Meteorological Information Center of the China Meteor­ological Administration. It includes daily maximum, minimum, and average temper­ature data from 2,481 national-level ground observation stations across the country. The establishment and production processes of this dataset have undergone rigorous quality control, and corrections have been made to existing errors to ensure a high degree of credibility. The specific station distribution is shown in Figure 1.

3.2 Interpolation Method

ADW is a method for interpolating discrete irregular station data into grid data in a certain space using a certain mathematical relationship. This method comprehensively considers the distance and azimuth relationship between stations and regular grid points and can achieve a better interpolation effect. Dunn^[10] created the HadEX3 dataset using this method. The specific formula is as follows: equation 1 refers to the grid point as the center, the correlation between the surrounding stations and the index value at this grid point presents an e-exponential decay, where x is the distance between the station and grid point, and x₀ is the search radius. In equation 2, w_i is the distance weight of each station and m is used to adjust the attenuation rate. In equation 3, the azimuth and distance are comprehensively considered, and W_i is the angular distance weight, where  is the azimuth of the station relative to the grid point and k is the number of stations within the search range. Since 1961, most meteorological stations have been relocated owing to changes in the underlying surface and surrounding environment, especially during the 20 years from 1961 to 1980, and the number of stations has increased rapidly. Therefore, the weight calculation of each station at each grid point was divided into five periods, 1961?C1965, 1966?C1970, 1971?C1975, 1976?C1980, and 1981?C2020, which not only considered the weight change but also made the interpolation calculation efficient. This study used the Python programming language to calculate equations 1?C3. First, the azimuth angle and distance between the station and the grid point are calculated through the custom function, that is, x in equation 1 and  in equation 3 are obtained, and then equations 1?C3 are written to establish the angular distance weighting interpolation method.

                                                                                                                (1)

                                                                                                                 (2)

                                 , ??                               (3)

3.3 Methodolgy

The establishment process of the Chinese extreme temperature index grid dataset is presented in Figure 2 and mainly includes four parts: pre-processing of the basic dataset, calculation of the 16 indices for the stations, interpolation to the latitude and longitude grid points, and output of the file.

(1) First, Python is used to perform data preprocessing on the Dataset of daily values of basic meteorological elements of national surface meteorological stations in China (V3.0). This included the positioning of station latitudes and longitudes, and conversion of data units, resulting in daily temperature data for 2481 stations in China from 1961 to 2020.

(2) Among the 16 extreme temperature indices, the calculation of TX90p, TN90p, TX10p, TN10p, WSDI, and CSDI, a total of six indices, required their relative thresholds to be determined. This was achieved using the percentile threshold method to calculate the climatological state thresholds for the period of 1961?C1990, while also excluding stations with more than 5% missing data. Subsequently, Python was utilized to calculate the 16 extreme temperature indices and obtain the calculation results for the stations. Considering the specific definitions of these indices, the temporal resolutions of the GSL, WSDI, and CSDI are only yearly, not monthly.

(3) ADW was used to interpolate and obtain the calculation results of the stations to the latitude and longitude grid points.

(4) According to the geographical scope of China, data masking was conducted, retaining only the data of the Chinese region (excluding territorial waters) and replacing the interpolation results of the remaining areas with missing measurement values. Python was used to input the dataset information (dataset name, creator, spatiotemporal resolution, etc.) and index information (index definition, units, etc.) for each file, specifying the data accuracy and missing data markers. Finally, the file was saved to generate the SimmEX dataset.

Figure 2  Flowchart of the technical process of SimmEX production

4 Data Results and Validation

4.1 Data Composition

The compressed package contained two folders, ANN and MON, storing 16 annual and 13 monthly indices, respectively. Each file was in .NetCDF format and named SimmEX_xx_ANN or SimmEX_xx_MON, where xx is the index code (Table 3).

Table 3 Dataset compositionand description

4.2 Data Results and Validation

Figure 3 takes the four indices of TXx, TX90p, SU, and WSDI as examples to show the results comparing the SimmEX and HadEX3 datasets of the Hadley Center in the United Kingdom. The spatial resolution of SimmEX is 0.25????0.25??, whereas that of HadEX3 is 1.25????1.875??. As can be seen from the spatial distribution of the climate states of these four indices, the TXx values (Figure 3a and 3b) of the two datasets were pathologically presented as low regions in Qinghai, Tibet, and western Sichuan and high regions in other regions. The TX90p values in the SimmEX dataset (Figure 3c) were smaller overall than those in the HadEX3 dataset (Figure 3d). SU (Figure 3e and 3f) showed low values in South China and

Figure 3  Climatological distribution maps (1961?C1990) of TXn (a), TX10p (b), TR (c), and CSDI (d) in SimmEX and HadEX3

high values in North China, while WSDI (Figure 3g and 3h) showed low values in Northeast China and high values in West China. It should be noted that the SimmEX dataset has a higher spatial resolution; therefore, it reflects more refined distribution characteristics. For example, the SimmEX WSDI presented high-value centers in small areas, such as Xinjiang, Tibet, Sichuan, and East China. However, the lower-resolution HadEX3 dataset could not reflect the fine-grained structural features at this small scale.

Figure 4 shows the interannual changes in TXx, TX90p, SU, and WSDI for the SimmEX and HadEX3 datasets. The comparison indicated that the four indices of the two datasets, in terms of interannual changes, the year with high and low values, and the changing trend, corresponded relatively well, with good consistency, especially TXx, TX90p, and WSDI (Figure 4a, b, and d), and the difference between the two sets of data is small. The SU index of the two datasets (Figure 4c) was the most different; SimmEX was lower than HadEX3 overall, but the interannual variation was consistent. In addition, the TNx, TNn, SU, TR, DTR, GSL, and other indices of the two datasets were also high or low, and there were certain differences that are not shown. In general, the indices of the two datasets were the same in terms of interannual variation, but there were different degrees of global deviation in the different indices. Although SimmEX and HadEX3 use the same definition for calculations and the same spatial interpolation method, SimmEX uses more original station data and denser spatial coverage, which is the main reason for the differences between the two datasets.

Figure 4  Interannual variations in TXx (a), TX90p (b), SU (c), and WSDI (d) in SimmEX and HadEX3

Figures 3 and 4 compare and analyze the spatial distribution and interannual variation differences of various indices in the SimmEX and HadEX3 datasets. We now scientifically evaluate the quality of the two datasets using statistics. The specific practices are as follows: each index of the two datasets was interpolated to 2,481 stations, and the root-mean-square error (RMSE) was used as the objective evaluation standard. The formula is given in Equation 4, where  is the value of the data in the dataset interpolated to the stations,  is the observed value of the stations, and  is the number of files.

                                                                                       (4)

Figures 5a and 5b show the evaluation results of the middle-age index data in the ANN folder. The figure shows that the average RMSE values of TXx, TXn, TNx, and TNn in SimmEX were approximately 2 ?? lower than that of HadEX3. The average RMSE of TX90p, TX10p, TN90p, and TN10p in SimmEX were approximately 2% lower than those in HadEX3. The average RMSE values of the five indices in SimmEX, including ID, FD, SU, TR, and GSL, were 10?C25 d lower than those in HadEX3. The average RMSE values of WSDI and CSDI in SimmEX were 1?C2 days lower than those in HadEX3. Additionally, the RMSE volatility of each index in SimmEX was smaller than that in HadEX3, indicating that it was more stable than HadEX3.

Figure 5c presents the evaluation results of the monthly index data for the MON folder. The figure shows that the average RMSE values of TXx, TXn, TNx, and TNn in SimmEX were approximately 2 ?? lower than those in HadEX3. The average RMSE levels of TX90p, TX10p, TN90p, and TN10p in SimmEX were 2%?C4% lower than those in HadEX3. The average RMSE levels of FD, SU, and TR in SimmEX were 1?C2 days lower than those in HadEX3. The RMSE of each index in SimmEX was less volatile than that in HadEX3. Each index in the SimmEX dataset outperformed that in the HadEX3 dataset.

Figure 5  RMSE of each index in SimmEX and HadEX3, where (a) and (b) are annual files and (c) presents monthly files

5 Discussion and Conclusion

Based on the dataset of daily values of basic meteorological elements of national surface meteorological stations in China (V3.0) and the Python programming language, 16 extreme temperature indices from 2,481 stations in China were calculated according to the international standard definition provided by ETCCDI. A gridded SimmEX extreme temperature index dataset was generated by interpolating the station data to longitude and latitude grid points using the ADW interpolation method. Through evaluation, it was found that the dataset quality was greatly improved compared with HadEX3. However, the following issues should be noted when using this dataset. Owing to the small number of stations in Western China, their distribution was sparser than that in Eastern China, so the reliability of data for Eastern China was higher than that of data for Western China. In particular, in the western Tibetan Plateau with sparse stations and complex topography, the data in this region were extrapolated and their reliability should to be considered; therefore, it is recommended to use it as appropriate. In the future, based on station data, we will integrate satellite remote sensing, reanalysis, and other multi-source data as basic data to compensate for the poor quality of this dataset in areas with sparse stations, such as western China.

Author Contributions

Chen, Q. Y. completed the calculation and processing of the dataset, and wrote the paper; Zhang, Y. provided the overall idea of dataset development and reviewed and revised the paper; Liu, X. Y. wrote the Python script of angular distance weighting method and corrected the paper. Lian, Q. L. helped improve the calculation script, to improve the efficiency of index calculation; Xu, J. J. provided the idea for the construction of the dataset.

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]      IPCC. Climate Change 2021: the Physical Science Basis [M/OL]. Cambridge: Cambridge University Press, 2021 [2024-05-08]. https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_Full_

Report.pdf.

[2]       Wu, S. H., Yin, Y. H. Impacts of climate extremes on human systems [J]. Climate Change Research, 2012, 8(2): 99-102.

[3]       Piao, S. L., Zhang, X. P., Chen, A. P., et al. The impacts of climate extremes on the terrestrial carbon cycle: A review [J]. Science China Earth Sciences, 2019(49): 1321-1334.

[4]       Wang, X. L., Hou, X. Y. Raster dataset of extreme temperature in the coastal area of China [J]. Journal of Global Change Data & Discovery, 2019, 3(1): 54-58.

[5]       Ma, W. D., Liu, F. G., Zhou, Q., et al. Development of extreme precipitation dataset of Qinghai-Tibet Plateau (1961?C2017) [J]. Journal of Global Change Data & Discovery, 2021, 5(1): 67-72.

[6]       Zhou, Q., Zhang, H. N., Ren, Y. X. Methodology of dataset development on extreme precipitation indexes in Weihe River Basin (1961?C2016) [J]. Journal of Global Change Data & Discovery, 2021, 5(1): 62-66.

[7]       Caesar, J., Alexander, L., Vose, R. Large-scale changes in observed daily maximum and minimum temperatures: Creation and analysis of a new gridded data set [J]. Journal of Geophysical Research: Atmospheres, 2006, 111(D5): D05101.

[8]       Chen, Q. Y., Zhang, Y., Liu, X. Y., et al. Grid dataset of extreme temperature index in China (1961?C2020) (V1.0) [J/DB/OL]. Digital Journal of Global Change Data Repository, 2024. https://doi.org/10.3974/geodb.2024.05.06.V1. https://cstr.escience.org.cn/CSTR:20146.11.2024.05.06.V1.

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

[10]    Dunn, R. J. H., Alexander, L. V., Donat, M. G., et al. Development of an updated global land in situ-based dataset of temperature and precipitation extremes: HadEX3 [J]. Journal of Geophysical Research: Atmospheres, 2020, 125(16): e2019JD032263.