the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
09 Dec 2021
09 Dec 2021
GPRChinaTemp1km: a high-resolution monthly air temperature dataset for China (1951–2020) based on machine learning
- 1Academy of Disaster Reduction and Emergency Management, Beijing Normal University, 100875 Beijing, China
- 2Faculty of Geographical Science, Beijing Normal University, 100875 Beijing, China
- 3The School of National Safety and Emergency Management, Beijing Normal University, 100875 Beijing, China
- 1Academy of Disaster Reduction and Emergency Management, Beijing Normal University, 100875 Beijing, China
- 2Faculty of Geographical Science, Beijing Normal University, 100875 Beijing, China
- 3The School of National Safety and Emergency Management, Beijing Normal University, 100875 Beijing, China
Abstract. An accurate spatially continuous air temperature dataset is crucial for multiple applications in environmental and ecological sciences. Existing spatial interpolation methods have relatively low accuracy and the resolution of available long-term gridded products of air temperature for China is coarse. Point observations from meteorological stations can provide long-term air temperature data series but cannot represent spatially continuous information. Here, we devised a method for spatial interpolation of air temperature data from meteorological stations based on powerful machine learning tools. First, to determine the optimal method for interpolation of air temperature data, we employed three machine learning models: random forest, support vector machine, and Gaussian process regression. Comparison of the mean absolute error, root mean square error, coefficient of determination, and residuals revealed that Gaussian process regression had high accuracy and clearly outperformed the other two models regarding interpolation of monthly maximum, minimum, and mean air temperatures. The machine learning methods were compared with three traditional methods used frequently for spatial interpolation: inverse distance weighting, ordinary kriging, and ANUSPLIN. Results showed that the Gaussian process regression model had higher accuracy and greater robustness than the traditional methods regarding interpolation of monthly maximum, minimum, and mean air temperatures in each month. Comparison with the TerraClimate, FLDAS, and ERA5 datasets revealed that the accuracy of the temperature data generated using the Gaussian process regression model was higher. Finally, using the Gaussian process regression method, we produced a long-term (January 1951 to December 2020) gridded monthly air temperature dataset with 1 km resolution and high accuracy for China, which we named GPRChinaTemp1km. The dataset consists of three variables: monthly mean air temperature, monthly maximum air temperature, and monthly minimum air temperature. The obtained GPRChinaTemp1km data were used to analyse the spatiotemporal variations of air temperature using Theil–Sen median trend analysis in combination with the Mann–Kendall test. It was found that the monthly mean and minimum air temperatures across China were characterized by a significant trend of increase in each month, whereas monthly maximum air temperature showed a more spatially heterogeneous pattern with significant increase, non-significant increase, and non-significant decrease. The GPRChinaTemp1km dataset is publicly available at https://doi.org/10.5281/zenodo.5112122 (He et al., 2021a) for monthly maximum air temperature, at https://doi.org/10.5281/zenodo.5111989 (He et al., 2021b) for monthly mean air temperature and at https://doi.org/10.5281/zenodo.5112232 (He et al., 2021c) for monthly minimum air temperature.
- Preprint
(3411 KB) -
Supplement
(14802 KB) - BibTeX
- EndNote
Qian He et al.
Status: open (until 26 Feb 2022)
-
CC1: 'Comment on essd-2021-442', QiangYu Lee, 10 Dec 2021
reply
Thank you for your work, you have provided valuable long-term monitoring data, and I have a few questions to ask for answers here, as follows:
1. Line 190, ask the machine learning dataset is
How to build it, specifically, what are the independent and dependent variables of the input model. Here I understand that you extracted the independent variables (i.e. latitude and longitude and elevation) related to the location of the corresponding meteorological stations in the training set section, and then extended them to all the space after you trained the model. Is the machine learning model trained every month or is it a model that covers the entire study period?
2. Can we open-source the core machine learning code, specifically the model building and training section, so as to deepen our understanding of the article method?-
AC1: 'Reply on CC1', Ming Wang, 11 Dec 2021
reply
Response1: Thank you for asking the question. The independent variables are latitude, longitude and elevation; the dependent variable is the air temperature (maximum temperature/mean temperature/minimum temperature). The machine learning models were trained for each month. We have explained the model training in the paper (See Preprint Lines 191-194): “We extracted the independent variables (i.e., latitude, longitude, and elevation) relating to the meteorological stations and randomly divided the processed data into a set for model training (70%) and a set for model evaluation and validation (30%). When train the models, the 10-fold cross-validation was used. We constructed a model for each month separately which means we have 840 models for the 840 months from 1951 to 2020. All algorithms were implemented in MATLAB R2020b.”
Response 2: Thanks for your suggestion. We agree that open source code can help to understand the work. However, due to the requirements of the foundation, we may make the codes available after the acceptance of the work.
-
AC1: 'Reply on CC1', Ming Wang, 11 Dec 2021
reply
-
RC1: 'Comment on essd-2021-442', Anonymous Referee #1, 19 Jan 2022
reply
The manuscript aims to produce a long term dataset of monthly 2m temperature over China at high spatial resolution on a 1x1 km grid. While the objective is appealing due to the challenges related to the complex topography and the irregular data availability in the target region, the applied methods show up with significant issues. The major issues are listed subsequently:
[1] The introduction discusses advantages and disadvantages of different information sources for the targeted dataset. While strong arguments for point-wise observational data are presented, long term reanalysis data products are not considered despite they provide consistent and spatio-temporally coherent information on the atmospheric state. It is unclear why such data is not considered to provide predictor variables.
[2] The method of data splitting leads to strong autocorrelation between the training and test dataset. Due to the spatial proximity of stations in both dataset, a fundamental requirement is hurt, that is the independency (or at least a minimization of dependency) between the training and test dataset. This is especially true for the stations located in the flat eastern parts of China with a dense observational network. Thus, the statistical model are prone to learn nearest neighbor-relations rather than learning real abstractions from the features, see, e.g. Kleinert et al., 2021 for a more detailed discussion on the requirement of splitting the test and training data temporally when stations are located close to each other.
[3] Only static features are used as predictors which implies that a model must trained for each month (!) of the period under consideration. Thus, dynamic information on the atmospheric state can exclusively deduced from the optimization procedure on the predictand. It is strongly recommended to introduce dynamical data as a predictor variable instead. Besides, the chosen predictors have periods with neglectable correlation with respect to the target quantity and important features such as the ambient topography (is the meteorological station located in a valley) is absent (see, e.g., Sha et al., 2020).
[4] The evaluation does not serve the objectives of the study. The stations in the test dataset are dominated by stations over flat terrain with a dense observational network. Thus, potential deficiencies in capturing the variations due to underlying complex topography are hidden. Indeed, Figure 5 indicates that residuals are considerably larger over the mountainous region.
[5] Several issues in the follow-up study are present such as (a) a focus on large-scale temperature patterns instead of fine-scale patterns in Section 4.2. to reason the high spatial resolution of the dataset, (b) the interpretation of patterns in the Xinjiang region which look like artefacts (bulls-eye pattern in winter months) and (c) the missing notification on the better performance of the reference method ANUSPLIN for July-months in the 70s, 80s and 90s.
[6] The comparison to the competing datasets ERA5 and FLADS is misleading due to the much coarser spatial resolution of these two datasets. A fair comparison would consult datasets with similar spatial resolution such as the dataset described in Peng et al., 2019.
Further minor issues are:
* Splitting into three distinct datasets is unnecessary. Rather merge it to one dataset with one DOI.
* Refer to statistical and dynamical downscaling techniques in the introduction.
* Provide references to the problems related to remote sensing data (see l.66).
* Describe the remapping of the STRM DEM data onto the 1x1 km grid (should be an averaging method).
* The used software tool MATLAB should be only mentioned once rather than being repeated three times. More details on the respective ML-technique would be appreciated.
* l.149: Should be 'ensemble machine learning'
* l.219: "Unnecessary reference to Equation 4 which directly follows the sentence.
* l.343f. This is sentence is barely comprehensible.Mentioned literature references:
- Kleinert, Felix, Martin G. Schultz, and Lukas H. Leufen. "IntelliO3-ts v1. 0: A neural network approach to predict near-surface ozone concentrations in Germany." Geoscientic model development discussions 2020. FZJ-2020-05012 (2020): 1-69
- Sha, Yingkai, et al. "Deep-learning-based gridded downscaling of surface meteorological variables in complex terrain. Part I: Daily maximum and minimum 2-m temperature." Journal of Applied Meteorology and Climatolog 59.12 (2020): 2057-2073.
- Peng, Shouzhang, et al. "1 km monthly temperature and precipitation dataset for China from 1901 to 2017." Earth System Data Science Data 11.4 (2019): 1931-1946
Qian He et al.
Data sets
GPRChinaTemp1km: 1 km monthly minimum air temperature for China from January 1951 to December 2020 He, Qian; Wang, Ming; Liu, Kai; Li, Kaiwen; Jiang, Ziyu https://doi.org/10.5281/zenodo.5112232
GPRChinaTemp1km: 1 km monthly mean air temperature for China from January 1951 to December 2020 He, Qian; Wang, Ming; Liu, Kai; Li, Kaiwen; Jiang, Ziyu https://doi.org/10.5281/zenodo.5111989
GPRChinaTemp1km: 1 km monthly maximum air temperature for China from January 1951 to December 2020 He, Qian; Wang, Ming; Liu, Kai; Li, Kaiwen; Jiang, Ziyu https://doi.org/10.5281/zenodo.5112122
Qian He et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 318 | 73 | 10 | 401 | 20 | 3 | 4 |
- HTML: 318
- PDF: 73
- XML: 10
- Total: 401
- Supplement: 20
- BibTeX: 3
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1