Based on the consideration of topographic information, a downscaling of the coarse resolution (30′) CRU dataset was carried out using a geo-weighted regression method. Temperature (Ta) and precipitation datasets were obtained for the Qilian Mountains from 1941-021, and the downscaled 5′ (1km) resolution dataset was evaluated using observations from meteorological stations. Compared with the original CRU dataset, the new dataset shows a 52.7% and 4.9% reduction in mean absolute error for Ta and precipitation, respectively. The root-mean-square errors for Ta and precipitation were reduced by 53.3% and 10.4%, respectively. The Nash-Sutcliffe efficiency coefficients for temperature and precipitation have improved by 25.4% and 9.5% respectively. The new dataset can provide detailed information for the analysis of climate change trends over multiple time scales. The dataset will help potential users to improve climate monitoring, modelling and environmental studies in the Qilian Mountains region.
1. Name of Data
Ta1941-2021month.nc
Pre1941-2021month.nc
2.Data description of attribute items
The data stored using the NetCDF format. Thus, each file contains 972 months of data and requires 4.69GB of storage space. Each file name indicates the data contained in the file, For example, the file named Ta1941-2021month. nc contains month temperature data from 1941 to 2021. The total number of NetCDF files is 2, and the total size of the dataset in nc format is approximately 9.38 GB.Temperature is measured in degrees Celsius and precipitation in millimeters. The spatial resolution of data is 1km.
(1) CRU datasets.The monthly mean temperature and precipitation were obtained for January 1941 to December 2021with a spatial resolution of 30′ from the CRU TS v4.02 dataset (http://www.cru.uea.ac.uk, last access: 24 June 2022).
(2) DEM data with the spatial resolution of 30 m were downloaded from http://www.gscloud.cn. The slope, aspect, and topographical relief layers were extracted by ArcGIS 10.8.2.
(3) Observations.
The observed long-term monthly temperature and precipitation data across Qilian Mountains were obtained from eight national meteorological stations (http://data.cma.cn/en) and 11 local meteorological stations.
1. Processing
First, the environmental variables at 0.5′ (1 km) and 30′ resolutions and the original CUR data at monthly scales were prepared.
Second, we selected variables that correlate with tetemperature and precipitation, named as explanatory variables.
Third, the explanatory variables and original CRU data were inputted into the GWR model, and the intercepts, residuals, and coefficients are obtained.
Fourth, the intercepts, residuals, and coefficients were interpolated to obtain 1 km raster layers. These layers were then combined with explanatory variables at 1 km resolution to develop temperature and precipitation data with a high resolution.
2.GWRmodel,
The regional regression model, was first proposed by Chris et al. (1996). This model has been widely applied in research about the dynamic and scale-dependent characteristics of the relationships between the dependent and the explanatory variables (Foody, 2003). The regression model is expressed as follows:
𝑌_𝑗=𝛽_0 (𝑢_𝑗,𝑣_𝑗 )+∑1_(𝑖=1)^𝑝▒〖𝛽_𝑖 (𝑢_𝑗,𝑣_𝑗 ) 𝑋_(𝑖,𝑗 ) 〗+𝜖_𝑗,
where (uj, vj), βo (uj, vj), βi (uj, vj), and εj are the geographical coordinates, intercept, slope (regression coefficient), and regression residual at the jth point, respectively; and p denotes the number of environmental variables. The intercept is the estimate of the local constant term, the slope is the local estimate of coefficient for each variable, and the residual represents the difference between the observed and the predicted values of the dependent variable. Yj represents the jth observation of the dependent variable; and Xi,j is the jth observation of the ith independent variable. The basic assumption of the GWR is that an observation being closer to the jth point has a higher influence on the local coefficient for the location, which acts as a distance decay function that depends on the distance from the jth point to its adjacent points.
Reference：Chris, Brunsdon, A, Stewart, Fotheringham, Martin, E, & Charlton. (1996). Geographically Weighted Regression: A Method for Exploring Spatial Nonstationarity. Geographical Analysis.
(3) GWR 4.0 software
GWR 4.0 software (http:// geoinformatics.wp.st-andrews.ac.uk/gwr) was used to establish the GWR model in our study.
Accuracy of Data：
Compared to the original CRU dataset, the new dataset shows a 52.7% and 4.9% reduction in mean absolute error for temperature and precipitation respectively. The root-mean-square errors for temperature and precipitation were reduced by 53.3% and 10.4%, respectively. The Nash-Sutcliffe efficiency coefficients for temperature and precipitation improved by 25.4% and 9.5%, respectively.
| # | number | name | type |
| 1 | 2019YFC0507404 | Monitoring of typical ecological environment changes in nature reserves | National key R & D plan |
| 2 | 2020-SF-146 | Comprehensive monitoring and functional assessment of typical ecosystems in Qilian Mountain National Park | other |
| 3 | 2019QZKK05010119 | Major special sub-projects of the Ministry of Science and Technology |
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.ncdc.NIEER.db2489.2022
10.12072/ncdc.NIEER.db2489.2022