The data set includes stable oxygen isotopes of precipitation in Chinese Mainland from 1870 to 2017. The time and spatial resolutions are monthly and 50-60 kilometer contour lines, respectively. Due to limited observational data and different time periods in climate model simulations, multiple methods and datasets were used to construct a long-term dataset.
(1) During the period of 1979-2001, CNN fusion method was used to fuse eight simulation results of five iGCMs (CAM2, GISS E, HadAM3, LMDZ4, and MIROC32) with observation results.
(2) Between 2002 and 2007, six simulation results of three iGCMs (GISS E, LMDZ4, and MIROC32) were fused using CNN fusion method.
(3) For the period from 1969 to 1978, the CNN fusion method was used to fuse four simulations of three iGCMs (CAM2, GISS E, and HadAM3).
(4) For the period of 1958-1968, two BCM were used to calibrate two iGCM simulations (CAM2 and HadAM3), and then the ensemble average (the average of four simulations) was calculated.
(5) For the periods of 1870-1957 and 2008-2017, two BCMs were used to modify an iGCM simulation (HadAM3 for 1870-1957 and LMDZ4 amplification for 2008-2017), and then the ensemble average (the average of the two simulations) was calculated.
When using this dataset, it should be noted that the δ 18Op generated by CNN fusion method during the period of 1979-2007 may be more reliable, as the performance comparison between BCMs and DFMs shows that CNN fusion method performs better than BCMs, and the number of iGCM simulated data fusion used during this period is greater than other periods.
The associated data for generating equivalent landscapes can be obtained as follows. GNIP data can be obtained from the International Atomic Energy Agency/World Meteorological Organization（ https://nucleus.iaea.org/wiser ）Obtain the GNIP database. The TNIP data is sourced from the National Qinghai Tibet Plateau Science Data Center (DOI: 10.11888/Geogra.tpdc.270940). IGCM data can be downloaded from SWING2 (available at https://data.giss.nasa.gov/swing2/ The LMDZ4 scaling data was provided by Dr. Camille Risi from Laboratoire de M é t é orologie Dynamique in France.
During the common period of 1969-2007, CNN fusion method was used to generate contour maps, and in other years, bias correction method was used to generate contour maps. The generated contour map displays a spatiotemporal distribution similar to the observed data, which is reliable and useful, and can provide strong support for tracking atmospheric and hydrological processes.
The generated isomorphic dataset has high spatiotemporal resolution and covers a long sequence from 1870 to 2017. Compared with existing iGCMs, contour lines have higher quality and stability for large regions of China on a monthly scale. Thanks to the characteristics of optimal neural networks and bias correction methods, contour maps fully utilize observational data and integrate the advantages of various iGCMs. That is to say, by combining data fusion and correction methods, all observation data and iGCM simulations were maximally utilized to ensure the highest accuracy throughout the entire period.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6459.2024
10.5281/zenodo.5703811