A Dataset on the Spatial Distribution of Annual
Precipitation in Middle and Western China (2010)

Jiang, Y. H.^1,2  Li, B. L.^1,2*  Yuan, Y. C.¹  Gao, X. Z.¹  Zhang, T.^1,2  Liu, Y.^1,2
Li, Y.^1,2  Li, H.³  Luo, Z. Y.³  Ma, Q.³  Wang, X. M.³  Ciren, D. J.⁴

1. State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographic Sci

ences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;
2. University of Chinese Academy of Sciences, Beijing 100049, China

3. Meteorological Bureau of Mongolian Autonomous County of Henan, Tibetan Autonomous Prefecture of Huangnan 811599, Qinghai, China;

4. Lhasa Meteorological Bureau, Lhasa 850000, Tibet, China

Abstract: The data set on the spatial distribution of annual precipitation in middle and western China (2010) is a simulation of regional precipitation with High Accuracy Surface Modeling (HASM), applying Hybrid Interpolation. The method includes three steps: First, steady precipitation changes within the space, or the trend surface, is represented with the TRMM data (spatial interpolation results from station observations values); then the residue values without the trend surface is calculated by incorporating the ground station observation values and interpolated with HASM, obtaining the residual field in the space with unsteady changes; finally, the interpolation calculation is completed by adding the trend surface and the residual field. The results show that HASM simulation with TRMM as the background field is significantly more precise and more adaptive than traditional interpolation methods. Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) are used as the precision assessment indexes for the results. On a global scale, MAE and RMSE are 125.15mm and 155.80mm, respectively; on a local scale, MAE and RMSE are 167.53mm and 228.81mm, respectively. An article about the findings from the data set has been published on the Journal of Geo-Information Science 2015 Vol. 17 Issue 8.

Keywords: precipitation; TRMM satellite; high accuracy surface modeling; middle and west China; Journal of Geo-Information Science

1 Introduction

Precipitation is an important environmental variable and plays an essential role in surface runoff, atmospheric movement, and agricultural resources. Large-scale precipitation data are usually based on the results of discrete site observations and are obtained using spatial interpolation. The traditional spatial interpolation method does not consider the characteristics of the spatial surface itself. In the modeling process, the constraint effect of the intrinsic factors on the surface reconstruction is neglected, while HASM could effectively resolve the issue of peak flattening and boundary turbulence that are difficult to avoid with traditional interpolation methods^[1]. However, in areas with fewer stations, the feasibility of the HASM model is greatly restricted. Satellite remote sensing data allows for large-scale and synchronous observation, which can make up for the limited observation range of ground stations. Therefore, the accuracy of spatial simulation of regional precipitation could be improved by integrating satellite precipitation product information while simulating precipitation with HASM^[2].

The data set on the spatial distribution of annual precipitation in middle and western China (2010)^[3] is funded by the Department of Science and Technology projects of the People??s Republic of China. This product is developed on the basis of the readily available TRMM 3B43 V7 data product with 0.25????0.25?? resolution^[4?C5]. To resolve the issue of poor HASM accuracy in regions with fewer stations, setting the TRMM data as the background field (trend surface) by using hybrid interpolation, the residual field is revised with HASM (after removing the trend) to improve HASM??s depicting capacity of spatial details of regional precipitation. The revised product is compared with the results generated from ground station interpolation to validate its effectiveness.

2 Metadata of Dataset

The metadata for the ??A dataset on the spatial distribution of annual precipitation in middle and western China (2010)?? is summarized in Table 1. It includes the name, authors, geog­ra­p­h­ical areas, decade of the data, temporal resolution, spatial resolution, composition of the data set, platform for data publication and sharing services, and data sharing policies. etc.

Table 1  Metadata summary for the ??A dataset on the spatial distribution of annual precipitation in middle and western China (2010)??

(To be Continued on the next page)

(Continued)

3 Methods

3.1 Algorithm Principles

Precipitation interpolation results are more uncertain in sparce stations and in areas with greater ground surface change. Remote sensing precipitation products could well manifest the spatial distribution patterns of precipitation. Therefore, this data set applies hybrid interpolation incorporated with the HASM model and remote sensing precipitation product to simulate the regional precipitation. The method includes three steps: First, steady precipitation changes within the space, or the trend surface, is represented with the TRMM data (or spatial interpolation results from contrast methods); then the residue values without the trend surface is calculated by incorporating the ground station observation values and interpolated with HASM, obtaining the residual field in the space with unsteady changes; finally, the final interpolation results are obtained by adding the trend surface and the residual field.

3.2 Technical Route

In the hybrid interpolation, HASM, IDW, Spline AND kriging interpolation results and TRMM data were applied to calculate the trend surface. The respective interpolation results are expressed as HASM_I, HASM_S, HASM_K and HASM_t. The HASM model equation system is resolved with iteration, using the Preconditioned Conjugate Gradient ^[7] (Figure 1).

Large-scale model simulation accur­a­cy is tested, using randomly selected standard weather stations as the mode­lling points and the remaining weather stations as global validation points; Without changing the modelling points, all local weather stations are used as local validation points to test the simulation effects of small-scale models. HASM of the background field is simulated and its results are compared by using the TRMM data product, IDW, Spline and Kriging interpolation results with 0.25????0.25?? as the analysis unit.

4 Results and Validation

4.1 Composition of Results

The data set is composed of 2 folders:

(1) Folder ??grids_precip?? stores 4 files in the tiff format: hasm_idw2010.tif, hasm_kriging2010.tif, hasm_spline2010.tif and hasm_trmm2010.tif, representing the 2010 precipitation data with Inverse Distance Weighted (IDW), Kriging interpolation, Spline interpolation and TRMM data results as the HASM drive field and using hybrid interpolation. The spatial resolution is 0.25??;

(2) Folder ??pts_precip?? stores 2 vector files in the .shp format, including stps96 and stps 25, which respectively stores the location of the 96 national weather stations and 25 local weather stations in the region of interest (field lat and lon refer to latitude and longitude respectively), station code (field no_st) and the accumulated precipitation in 2010 (field precip).

4.2 Data Results

It can be sen from Figure 2 that precipitation data from four sources have similar spatial distribution patterns. Results of HASM simulation with TRMM as the drive filed have manifested more information on spatial changes in precipitation in regions without station observation, which are superior to other HASM simulation results with drive fields generated from traditional interpolation methods. In Figure 2, there are few discrepencies in the HASM-simulated spatial patterns of precipitation in various background fields, generally exhibiting a gradual decrease in precipitation from the southeast to the northwest. Results of HASM simulation with TRMM as the drive filed have manifested more information on spatial changes in precipitation in regions without station observation, which are superior to other HASM simulation results with drive fields generated from traditional interpolation methods. Locus A in the figure is a drastic transition area from humid and semi-humid regions to semi-arid regions, where the precipitation should present a rapidly decreasing trend from the southeast to the northwest ^[8]. Of the four methods, only the HASM_T method was able to simulate such changes. Locus B is situated in the Weihe Plain, where HASM_T was able to simulate precipitation distribution that is consistent with the results of previous studies. Locus C is the upstream of the Lishui River on the northen branch of the Wuling Mountains. It is one of the precipitation centres of Hunan Province. Locus D is the Hengyang-Shaoyang hugelland region and a primary ??arid area?? ^[9] of Hunan Province. HASM_T dinstinctly reflected these regions of precipitation extrema.

4.3 Validation

It can bee seen from Figure 2 that accuracy of HASM simulation with TRMM as the drive filed is superior to that of HASM simulation with traditional interpolation methods as the drive field. In the global validation results, the MAE and RMSE of HASM_T is 125 and 156mm (Table 2), respectively. The MAEs of HASM_I, HASM_S and HASM_K are 212, 234 and 192mm, which are respectively 70%, 87% and 54% higher than HASM_T results; the RMSEs are 260, 328 and 241mm, which are respectively 67%, 110% and 54% higher than HASM_T results. In the local validation results, the MAE and RMSE of HASM_T is 168 and 229mm, respectively. The MAEs of HASM_I, HASM_S and HASM_K are 196, 198 and 197mm, which are respectively 17%, 18% and 17% higher than HASM_T results; the RMSEs are 263, 260 and 256mm, which are respectively 15%, 14% and 12% higher than HASM_T results.

Figure 2  HASM simulated precipitation distribution in different background fields

Table 2  HASM calculation accuracy in different background fields (mm)

5 Discussion and Conclusion

This data set uses TRMM satellite precipitation data as the background field and revises the residual field by incorporating the HASM model. The spatial distribution of precipitation in 2010 is simulated with middle and western China as an example; in the meantime, precipitation spatial distribution data are obtained with traditional interpolation. HASM simulation with TRMM as the background field is significantly more accurate than traditional interpolation methods. On a global scale, MAE and RMSE are 125.15mm and 155.80mm, respectively; on a local scale, MAE and RMSE are 167.53mm and 228.81mm, respectively. HASM simulation results with local region TRMM as the background field can better reflect the basic spatial patterns of precipitation. The results can show the rapid changes in precipitation, reflect precipitation extrema areas and effectively avoid outstanding phenomenon due to high observation values of individual stations occurring in traditional spatial interpolation algorithms. In different sub-regions, HASM simulation with TRMM as the background field has lower deviation better adaptability.

Author Contributions

Li, B. L. and Jiang, Y. H. created the overall design for the data collection development; Liu, Y. and Li, Y. processed the TRMM precipitation data; Zhang, T. and Yuan, Y. C. designed the model and algorithm; Gao, X. Z. performed data verification; Li, H., Luo, Z. Y., Ma, Q., Wang, X. M. and CiRen, DuoJi. provided the weather station data. Jiang, Y. H. and Li, B. L. wrote the data paper.

References

[1]       Zhao, N., Jiao, Y. M., et al. Downscaling of TRMM Satellite Precipitation Data [J]. Journal of Geo-Information Science, 2018, 20(10): 20?C27.

[2]       Zhang, T. Li, B. L., Zhao, N., et al. Analysis on High Accuracy Surface Modeling in Regional Rainfall

[3]       Estimation Combined with TRMM Data [J]. Journal of Geo-Information Science, 2015, 17(8): 895?C901

[4]       Jiang Y. H., Li B. L., Yuan, Y. C., et al. A dataset on the spatial distribution of annual precipitation in middle and western China (2010) [DB/OL], Global Change Research Data Publishing and Re-pository, 2019. DOI: 10.3974/geodb.2019.05.18.V1.

[5]       Kummerow, C., Barnes, W. The tropical rainfall measuring mission (TRMM) sensor package [J]. Journal of Atmospheric and Oceanic Technology, 1998, 15: 809?C817.

[6]       Huffman, G. J., Bolvin, D. T., Nelkin, E. J., et al. The TRMM multisatellite precipitation analysis (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at fine scales [J]. Journal of Hydrometeorology, 2007, 8(1): 38?C55.

[7]       GCdataPR Editorial Office. GCdataPR data sharing policy [OL]. DOI: 10.3974/dp.policy.2014.05 (Updated 2017).

[8]       Golub, G. H., Van Loan, C. F. Matrix Computations [M]. Beijing: Posts & Telecom Press, 2009.

[9]       Zhao, X., Ma, J. Z., Nie, Z. Q., et al. Spatiotemporal pattern of precipitation in Tianshui region and its response to the Indian Ocean dipole [J]. Arid Zone Research, 2009, 26(6): 917-922

[10]    You, X., Li, C., Yu, W. H. Analysis of Precipitation Characteristics in Hunan Province [J]. Hunan Hydro & Power, 2004 (6):41-43.