1 km Grid Precipitation Dataset in the Three-river
Headwaters Region (2009–2013)
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 Luo, Z. Y.3 Li, H.3 Ma, Q.3 Wang, X. M.3 CiRen, DuoJi.4
1. State Key Laboratory of Resources and
Environmental Information Systems, Institute of Geographic Sciences 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: This dataset includes the downscaled Tropical
Rainfall Measuring Mission (TRMM) 3B43 precipitation based on rainfall observations
from rain gauges and related contrasting or auxiliary data. Downscaled TRMM data was based on a quadratic
parabolic profile (QPP) model. This method involves two steps. (1) QPP model
parameters estimation. In three-river headwaters region, the elevation of
maximum precipitation corresponds to the elevation of maximum NDVI. Thus, the
elevation of maximum precipitation could be determined based on the spatial
location of the peak NDVI. Subsequently, estimating the precipitation
at the elevation of maximum precipitation as well as the quadratic coefficients
in parabolic equation of precipitation. (2) Spatial extrapolation of
model parameters. The parameters at 0.25° resolution are
spatially extrapolated in inverse distance weighted interpolation. The
bilinear interpolation method is then employed to resample
the parameters from 0.25° to 1-km resolution to obtain the ultimate
parameters of the downscaled model for each pixel. The results show that downscaled
TRMM 3B43 data in QPP model are more accurate than those obtained in conventional
statistical downscaling methods. The average root-mean-square errors (RMSEs) and
mean absolute percent errors (MAPEs) calculated with national observation data
of May to September and growing season in 2009–2013
are 14, 18, 19, 13, 16 mm and 14%, 12%, 12%, 12%, and 17%, respectively. The
dataset is stored in the WGS84 coordinate system as vector data in a .shp
format and raster data in a Grid or tif format. A paper related to this dataset
has been published in Remote Sensing of Environment (volume 215, 2018).
Keywords: precipitation;
TRMM satellite; downscaling; Three-River Headwaters
region and its nearby regions; Remote Sensing of Environment
1 Introduction
Precipitation, an important environmental element,
plays a role not to be overlooked in areas such as surface runoff, atmospheric
motions, and agricultural resources. However, due to the randomness,
precipitation exhibits relatively complex temporal and spatial variation patterns.
It is difficult to extrapolate precipitation observation data acquired at the
limited number of ground stations, particularly in regions with relatively few
stations. Satellite remote sensing products cover large areas and consist of
repeated observations. As a consequence, satellite remote sensing has become an
important means for acquiring information on temporal and spatial variations of
precipitation. However, due to such factors as topography, precipitation in
mountainous regions exhibits notable heterogeneity. As a result, satellite
precipitation data for mountainous regions are of extremely high uncertainty
and unable to meet the actual requirements[1–3]. In view of this, a finer precipitation dataset was
developed in this study by spatial downscaling of precipitation data retrieved
from satellite remote sensing data.
1 km Grid Precipitation Dataset in the
Three-river Headwaters Region (2009–2013)[4]
was supported by Ministry of Science and Technology of the People´s Republic of
China and was produced using the widely available TRMM 3B43 V7 data product
(0.25°×0.25°)[5, 6]. The
TRMM data product in three-river headwaters region is of relatively low
accuracy, and to solve the problem, assuming that precipitation is jointly determined
by macroscopic geographical factors and local elevations, and there is a strong
correlation between normalized difference vegetation index (NDVI) and
precipitation in the region. Under these assumptions, a quadratic parabolic
profile (QPP) model was employed to downscale the TRMM data. The parameters of
the relationship between NDVI and digital elevation model (DEM) were determined
based on high-resolution NDVI data, then to estimate the parameters between DEM
and precipitation. Finally, downscaled TRMM precipitation data was based on the
high-resolution DEM data[3]. The
revised product is compared with the results generated from the ground site
interpolation method for validation.
2 Metadata of Dataset
The metadata for
the “1 km Grid Precipitation Dataset in the Three-river Headwaters Region
(2009–2013)”[4] is summarized in
Table 1. It includes the full name, short name, authors, year, temporal
resolution, spatial resolution, data format, data size, data files, data
publisher, data sharing policy, etc.
3 Method
3.1 Algorithm principle
According
to the theory of precipitation in mountainous established by Fu[8],
precipitation in a mountainous region is jointly determined by precipitation
affected by macroscopic geographic factors and precipitation variation resulting
from the difference in local elevation, and can thus be represented and
calculated by a parabolic equation. The elevation difference of region is
large, and precipitation in this area is greatly affected by the terrain, which
satisfies the application condition of parabolic equation. In the studying
region, the annual average precipitation is less than 1,000 mm, and
precipitation has a positive correlation wtih NDVI in such region when
excluding the effects of local topography. Therefore, the same function should
fit the relationships between elevation and precipitation or NDVI. When not
taking into consideration the effects of local topography, there is a positive
correlation between precipitation and vegetation growth in sub-humid and
semiarid regions, which can be represented by the linear response relationship
between precipitation and NDVI[9, 10]. Hence,
algorithm used in this study is to assume that both precipitation and NDVI are
in a quadratic parabolic relationship with elevation. The parameters of the NDVI–DEM function
Table 1 Metadata summary of “1 km grid precipitation dataset in the
Three-river headwaters region (2009–2013)”
Item
|
Description
|
Dataset full name
|
1 km grid precipitation dataset in the Three-river
headwaters region (2009–2013)
|
Dataset short name
|
PrecipThreeRiverHeadwaters_2009-2013
|
Authors
|
Jiang, Y. H. N-8765-2019, Institute of Geographic Sciences
and Natural Resources Research, ChineseAcademy of Sciences,
jiangyh@lreis.ac.cn
Yuan, Y.C. N-9047-2019, Institute of Geographic Sciences
and Natural Resources Research, Chinese Academy of Sciences,
yuanyc@lreis.ac.cn
Gao, X, Z. N-1655-2019, Institute of Geographic Sciences
and Natural Resources Research, Chinese Academy of Sciences,
gaoxz@lreis.ac.cn
Zhang, T. N-8690-2019, Institute of Geographic Sciences
and Natural Resources Research, Chinese Academy of Sciences, zhangtao@lreis.ac.cn
Liu, Y. N-8844-2019, Institute of Geographic Sciences and
Natural Resources Research, Chinese Academy of Sciences,
liuy.18b@igsnrr.an.c.cn
Li, Y. Y-4384-2019, Institute of Geographic Sciences and
Natural Resources Research, Chinese Academy of Sciences, liying9391@126.com
Luo, Z. Y., Meteorological Bureau of Mongolian Autonomous
County of Henan, Tibetan Autonomous Prefecture of Huangnan, 393352158@qq.com
Li, H., Meteorological Bureau of Mongolian Autonomous
County of Henan, Tibetan Autonomous Prefecture of Huangnan, lh691208@163.com
Ma, Q., Meteorological Bureau of Mongolian Autonomous
County of Henan, Tibetan Autonomous Prefecture of Huangnan, 107082968@qq.com
Wang, X. M., Meteorological Bureau of Mongolian Autonomous
County of Henan, Tibetan Autonomous Prefecture of Huangnan, 2444869807@qq.com
Ciren, D. J., Lhasa Meteorological Bureau,
LSNSE111@126.com
|
Geographical region
|
31.65–36.27°N, 89.40–102.38°E, including
16 counties and one town in central and southern Qinghai: Xinghai county,
Zekog county, Henan county, Gade county, Maqin county, Banma county, Yushu
Tibetan autonomous prefecture, Chindu county, Zadoi county, Zhidoi county,
Madoi county, Qumarleb county, Nangqen county, Dari county, Jiuzhi county,
Tongde county, Tanggula town
|
Year
|
2009–2013 Temporal
resolution 1 month Spatial resolution 1 km
|
Data format
|
.shp, .tif, .grid
Data size 344 MB (compressed)
|
Dataset
files
|
The dataset consists of four folders.
(1) The “QPR_Precip” folder contains 5.Grid files:
QPR_2009、QPR_2010、QPR_2011、QPR_2012
and QPR_2013 are downscaled TRMM 3B43 cumulative precipitation (spatial
resolution: 1 km; unit: mm) in QPP model for May, June, July, August,
September, and the growing season of 2009, 2010, 2011, 2012 and 2013,
respectively.
(2) Control_Precip folder contains 15.Drid files:
ER_2009、ER_2010、ER_2011、ER_2012
and ER_2013 are downscaled TRMM 3B43 cumulative precipitation (spatial
resolution: 1 km; unit: mm) in for exponential regression (ER) model May,
June, July, August, September, and the growing season of 2009, 2010, 2011,
2012 and 2013, respectively.
MLR_2009、MLR _2010、MLR
_2011、MLR _2012 and MLR _2013 are downscaled TRMM 3B43
cumulative precipitation (spatial resolution: 1 km; unit: mm) in multiple
linear regression (MLR) model for May, June, July, August, September, and the
growing season of 2009, 2010, 2011, 2012 and 2013, respectively.
GWR_2009、GWR_2010、GWR_2011、GWR_2012
and GWR_2013 are downscaled TRMM 3B43 cumulative precipitation (spatial
resolution: 1 km; unit: mm) in geographically weighted regression (GWR) model
for May, June, July, August, September, and the growing season of 2009, 2010,
2011, 2012 and 2013, respectively.
(3) DEM file contains a .Grid file named dem1km, which is
the information on elevation variation within the study area (spatial
resolution: 1 km).
|
|
(4) NDVI folder contains five .grid files named
ndv1km_2009, ndv1km_2010, ndv1km_2011, ndv1km_2012, and ndv1km_2013, which
are vegetation growth data (spatial resolution: 1 km) for the growing season
of 2009, 2010, 2011, 2012, and 2013, respectively.
|
Foundations
|
Ministry of Science and
Technology of P. R. China (2016YFC0500205, 2015CB954103)
|
Computing Enviroment
|
Python 2.7; ArcGIS campus license of
Institute of Geographical Sciences and Natural Resources Research, Chinese
Academy of Sciences, Beijing, China.
|
Data publisher
|
Global Change Research Data Publishing & Repository,
http://www.geodoi.ac.cn
|
|
|
|
(To
be continued on the next page)
(Continued)
Item
|
Description
|
Address
|
No. 11A, Datun Road, Chaoyang District, Beijing 100101, China
|
Data sharing policy
|
Data from the Global Change Research Data Publishing & Repository
includes metadata, datasets (data products), and publications (in this case,
in the Journal ofGlobal Change Data & Discovery). Datasharing policy includes:
(1) Data are openly
available and can be free downloaded via the Internet; (2) End users are
encouraged to use Data subject
to citation; (3) Users,who are by definition also value-added service
providers, are welcome to redistribute Data
subject to written permission from the GCdataPR Editorial Office and
the issuance of a Data redistribution
license; and (4) If Data are
used to compile new datasets, the ‘ten percent principal’ should be followed
such that Data records
utilized should not surpass 10% of the new dataset contents, while sources
should be clearly noted in suitable places in the new dataset[6]
|
Communication and searchable system
|
DOI, DCI, CSCD, WDS/ISC, GEOSS, China GEOSS
|

Figure 1 Technical route for the dataset
development
|
are first determined based on
high-resolution NDVI and DEM data, then to
estimate the parameters between DEM and precipitation. Finally, downscaled TRMM precipitation data was
based on the high-resolution DEM data[3].
3.2 Technical route
The dataset was
produced in TRMM data downscaling method that accounts for macroscopic geographic
factors and local elevations[3].
This method involves the following two main
steps (Figure 1): (1)
Estimation of the QPP model parameters. Based on the assumption in the algorithm
principle that there are similar parabolic relationships between precipitation
and elevation or NDVI, the maximum precipitation should be at the location
where the peak NDVI occurs[9–10].
(2) Spatial extrapolation of model parameters. The parameters at 0.25°
resolution are spatially extrapolated in inverse distance weighted
interpolation. The bilinear interpolation method is then employed to resample
the parameters from 0.25° to 1-km resolution to obtain the ultimate parameters of the downscaled model for each pixel.
4 Data Results and Validation
4.1 Data Products
The dataset is
composed of 4 folders (QPR_Precip, Control_Precip, DEM and NDVI) as shown in Table
1.
4.2 Data results
DEM and NDVI were shown in Figure
2 and Figure 3. Precipitation data of July and growing
season in the region between 2009–2013 were
shown in Figures 4 and 5.
As shown in Figure 2, the elevation in three-river
headwaters region gradually increases from southeast
to northwest. A decrease trend of growing season NDVI can be observed in Figure
3 from southeast to northwest.

Figure 2 DEM for and spatial
distribution of national meteorological stations in the Three-river headwaters
region

Figure 3 Spatial distribution of NDVI in
the three-river headwaters region in the growing season of 2012
In Figures 4 and 5,
precipitation of July and the whole growing season in this region mainly
gradually decreases from southeast to northwest, which corresponds to the
topography and the direction of water vapor (southeast monsoon) in three-river
headwaters region. Locally, the maximum precipitation often occurs on mountain
slopes instead of valleys, and this corresponds to the theory of maximum
precipitation elevations.
4.3 Validation of data results
Mean Absolute Error (MAE)
and Mean Absolute Percent Error (MAPE) are the accuracy
evaluation indicator to test the accuracy of the downscaled TRMM monthly and
cumulative precipitation data for the growing season in the period 2009–2013. The downscaled data are of relatively high accuracy and
can relatively satisfactorily meet the accuracy requirements of relevant research for the spatial and temporal
distribution patterns of precipitation (Table 2).
Table 2 Average RMSEs and
MAPEs between the downscaled data and observation data acquired at national
stations for the growing season in the period 2009–2013[3]
Index
|
May
|
June
|
July
|
August
|
September
|
Cumulative
|
RMSE (mm)
|
14
|
18
|
19
|
13
|
16
|
62
|
MAPE (%)
|
14
|
12
|
12
|
12
|
17
|
11
|
5 Discussion and Conclusion
The dataset generation algorithm in this study
primarily takes into account the mechanism of formation of precipitation in
mountainous regions[8]. Based on
the theory of maximum precipitation elevations and the linear response relationship
between NDVI and precipitation, the TRMM precipitation data for three-river
headwaters region were downscaled to improve their accuracy and provide a more
accurate data product for analyzing the temporal and

Figure
4 Spatial distribution of precipitation in the three-river headwaters
region in July in the period
2009–2013

Figure 5 Spatial distribution of precipitation in the three-river
headwaters region in the growing season in the period 2009–2013
spatial distribution
patterns of precipitation. A notable decrease is found in the simulated regional
precipitation from southeast to northwest, which corresponds to the topography
and
the direction of
water vapor (southeast monsoon) in three-river headwaters region. Locally, the
maximum precipitation often occurs on mountain slopes instead of valleys, and
this corresponds to the theory of maximum precipitation elevations. The errors
of the downscaled data are also within a reasonable range, and these data can
meet the requirements of regional hydrological research.
The shortcomings of the algorithm in this study were not
take into consideration the topographical and geomorphic factors of three-river
headwaters region, and also neglected solid precipitation. Potential impact includes:
(1) Ignoring the different effects of terrain uplift on precipitation on the
windward and leeward slopes; (2) This algorithm is only applicable in
downscaling of growing-season precipitation, and was not for solid
precipitation in winter, thus was unable to generate precipitation data for the
whole year. Additionally, the accuracy of downscaled data is significantly
affected by the accuracy of the original product. If the pre-downscaling
precipitation data contain large errors, the accuracy of the downscaled data
product will remain inadequate. Therefore, this algorithm will give rise to
significant uncertainties when used in application research that requires
relatively high accuracy in absolute precipitation.
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; Luo, Z. Y., Li, H., Ma, Q., Wang, X. M. and Ci, ren, Duoji. provided
the weather station data.
Jiang, Y. H. and Li, B. L. wrote the data paper.
References
[1]
Hao, Z. C., Tong, K., Zhang, L.
L., et al. Applicability
Analysis of TRMM Precipitation Estimates in Tibetan Plateau [J]. Water, 2011, 31(5): 18–23.
[2]
Liu, Y. B., Fu, Q. N., Song, P., et al. Satellite Retrieval of
Precipitation: A Review[J]. Advance in Earth Sciences,
2011, 26(11): 1162–1172.
[3]
Zhang, T., Li, B., Yuan, Y., et al.
Spatial downscaling of TRMM precipitation data considering the impacts of
macro-geographical factors and local elevation in the Three-River Headwaters
Region [J]. Remote Sensing of Environment,
2018, 215: 109–127.
[4]
Jiang, Y. H., Li, B. L., Yuan, Y. C., et al. A 1-km grid precipitation
dataset for the Three-River Headwaters region and its nearby regions for the
period 2009–2013 [DB/OL], Global Change Research Data Publishing & Repository,
2019. DOI: 10.3974/geodb.2019.05.17.V1.
[5]
Kummerow, C., Barnes, W. The
tropical rainfall measuring mission (TRMM) sensor package [J]. Journal of Atmospheric and Oceanic
Technology, 1998, 15: 809–817.
[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):
38–55.
[7] GCdataPR Editorial Office. GCdataPR data
sharing policy [OL]. DOI: 10.3974/dp.policy.2014.05 (Updated 2017).
[8]
Fu, B. P. Mountain climate [M].
Beijing: Science Press, 1983.
[9]
Groeneveld DP, Baugh WM. Correcting satellite data to detect vegetation
signal for eco-hydrologic analyses[J]. Journal of
Hydrology (Amsterdam), 2007, 344(1-2): 135-145.
[10]
Chamaille-Jammes S, Fritz H, Murindagomo F. Spatial patterns of the
NDVI–rainfall relationship at the seasonal and interannual time scales in an
African savanna. International Journal of Remote Sensing[J], 2006, 27(23): 5185-5200.