Refined Evaluation of Satellite Precipitation Products against Rain Gauge Observations along the Sichuan–Tibet Railway

卫星降水产品在川藏铁路区域的精细化评估

+ Author Affiliations + Find other works by these authors
  • Corresponding author: Xiuping YAO, yaoxp@cma.gov.cn
  • Funds:

    Supported by the National Natural Science Foundation of China (42030611 and 42165005), the Second Tibetan Plateau Scientific Expedition and Research (STEP) Program (2019QZKK0103 and 2019QZKK0106), and Key Research and Development Plans of Tibet Autonomous Region in 2022 (XZ202201ZY0008G)

  • doi: 10.1007/s13351-022-1226-z

PDF

  • Being constructed in southwestern China, the Sichuan–Tibet Railway (STR) travels across the eastern Tibetan Plateau where there is the most complex terrain and changeable weather in the world. Due to sparse ground-based observations over the Tibetan Plateau, precipitation products retrieved by remote sensing are more widely used; however, satellite-based precipitation products (SPPs) have not yet been strictly and systematically evaluated along the STR. This study aims to evaluate the performance of six SPPs by a series of metrics with available ground observations along the STR during 1998–2020. The six SPPs include the datasets derived from the Tropical Rainfall Measuring Mission (TRMM), Climate Prediction Center morphing technique (CMORPH), Global Precipitation Measurement (GPM), Global Satellite Mapping of Precipitation (GSMaP), Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN), and Fengyun-2 satellites precipitation estimate (FY2PRE). The results indicate that most of the SPPs can capture the precipitation characteristics on multiple timescales (monthly, daily, hourly, and diurnal cycle) as shown by the evaluated metrics. The probability density functions of the daily and hourly precipitation are also well represented by the SPPs, and 30 mm day−1 and 16 mm h−1 are identified as the daily and hourly thresholds of extreme precipitation events along the STR. The best SPP varies at different timescales: GPM and GSMaP are suitable for the monthly and daily scale, and FY2PRE and GPM are suited to the hourly scale. In general, GPM is relatively optimum on multiple timescales, and PERSIANN gives the worst performance. In addition, the SPPs perform worse at higher altitudes and for more intense precipitation. Overall, the results from this study are expected to provide essential reference for using the SPPs in meteorological services and disaster prevention in the STR construction and its future operation.
    川藏铁路将穿越青藏高原东部,这是世界上地形最复杂的地区。复杂多变的降水及相关衍生地质灾害是川藏铁路建设和运营的主要挑战之一。由于当地的地面观测资料稀少,遥感降水产品可以提供至关重要的参考;然而,卫星降水产品(SPP)在川藏铁路沿线的准确性还没有得到很好的评估,这是在川藏铁路沿线使用SPP的瓶颈。本研究旨在评估六种广泛使用的SPP(包括TRMM、CMORPH、GPM、GSMaP、PERSIANN和FY2PRE)的性能;基于1998–2020年的川藏铁路沿线地面降水观测,通过一系列的定量指标给出了六种SPP在多个时间尺度上的表现和相对性能排名。结果表明,大部分SPP可以很好地捕捉到川藏铁路沿线降水的多时间尺度(月、日和小时)的特征和降水的日变化特征。不同时间尺度上表现最佳的SPP有所不同;相对而言,GPM在多个时间尺度的表现是最优的。此外,SPP在高海拔地区和降水强度较大的情况下表现较差。本文的评估结果为SPP应用于川藏铁路建设运营和灾害预防以及提高SPP在复杂地形地区的准确性,提供了重要的参考。
  • 加载中
  • Fig. 1.  (a) Distributions of the elevation (shading), automatic weather stations (AWSs; black dots), and national basic stations (red circles) over the eastern TP (red rectangles) along the Sichuan–Tibet Railway (STR; thick blue line). (b) The histogram (%) of the elevation of the rain gauges in the entire region along the STR and each sub-region, and the vertical red line denotes the average elevation of the region.

    Fig. 2.  Spatial variability of annual accumulated precipitation along the STR from (a) in situ rain gauges (the legend is the same as the color bar), (b) TRMM, (c) CMORPH, (d) GPM, (e) GSMaP, (f) PERSIANN, and (g) FY2PRE. The red thick line in each figure denotes the STR.

    Fig. 3.  Scatterplots of monthly precipitation from the gauge observation along the STR versus satellite-based precipitation products (SPPs) including (a) TRMM, (b) CMORPH, (c) GPM, (d) GSMaP, (e) PERSIANN, and (f) FY2PRE. The dash and solid lines denote best fitted performance (y = x) and linear fitting of the SPPs to observations, respectively.

    Fig. 4.  Taylor diagrams for correlation coefficients (CORR), standard deviation ratios (SDR), and normalized root-mean-square errors (RMSE) on the (a) entire ETP along the STR and (b) SCB, (c) WSC, (d) ET, and (e) CT.

    Fig. 5.  The two-dimensional distribution density of daily precipitation from the in situ rain gauges along the STR versus the SPPs of (a) TRMM, (b) CMORPH, (c) GPM, (d) GSMaP, (e) PERSIANN, and (f) FY2PRE.

    Fig. 6.  (a, c) Probability distribution function (PDF) and (b, d) cumulative distribution probability (CDF) of (a, b) daily precipitation days (PDs) and (c, d) total precipitation amount (TPA) with varying intensities over the entire ETP along the STR. The dashed lines denote the dominate precipitation intensities for the TPA (4 mm day−1) and the daily extreme precipitation (30 mm day−1).

    Fig. 7.  Spatial distributions of statistical metrics for the SPP daily precipitation of ≥ 4 mm day−1 along the STR: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The panels from top to bottom show the results of the TRMM, CMORPH, GPM, GSMaP, PERSIANN, and FY2PRE, respectively.

    Fig. 8.  The two-dimensional distribution density of hourly precipitation from in situ rain gauges along the STR versus SPPs: (a) CMORPH, (b) GPM, (c) GSMaP, (d) PERSIANN, and (e) FY2PRE.

    Fig. 9.  As in Fig. 6, but for the hourly precipitation (PHs: precipitation hours). The dash lines denote the precipitation intensities of 1 and 16 mm h−1.

    Fig. 10.  The spatial distributions of statistical metrics for the SPP hourly precipitation of ≥ 1 mm h−1 along the STR: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The panels from top to bottom show the results of CMORPH, GPM, GSMaP, PERSIANN, and FY2PRE, respectively.

    Fig. 11.  The diurnal cycle of precipitation from the AWS network and SPPs in the warm season (May–September) along the STR and in the sub-regions: (a) SCB, (b) WSC, (c) ET, and (d) CT. The gray shaded area denotes the hours in the nighttime.

    Fig. 12.  The elevation against the evaluation metrics of the daily precipitation: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The dash and solid lines indicate the linear regression of the metric to the elevation, and the dash (solid) line denotes that the metric decreases (increases) with increasing elevation.

    Fig. 13.  Scores of the (a) POD, (b) FAR, (c) CSI, and (d) ETS for SPPs with various daily precipitation intensities.

    Table 1.  Basic information about the ground-based rain gauges in the whole eastern Tibetan Plateau (ETP) along the STR, and its each sub-region. SCB indicates the Sichuan basin, WSC the western part of Sichuan, ET the eastern part of Tibet, CT the central part of Tibet, AWS the automatic weather station, and NBO the national basic observation

    RegionAverage altitude (m)Altitude range (m)AWS/NBO countAverage spatial density of rain gauges for
    AWS/NBO networks (No. per 100 km2)
    ETP2131 284–55802678/610.303/0.0070
    SCB 985 284–28511551/110.785/0.0072
    WSC3394 515–4536 695/220.289/0.0088
    ET40251618–5193 197/150.105/0.0075
    CT43483294–5580 235/130.120/0.0085
    Download: Download as CSV

    Table 2.  Satellite-based precipitation products. The full names of the institutes and datasets are detailed in Section 2.3

    DatasetInstitutionPeriodResolutionScaleSensorReference
    TRMMNASA1998.1–2019.12 0.25° × 0.25°3 hMultipleHuffman et al. (2007)
    CMORPHCPC/NOAA1998.1–2020.120.25° × 0.25°30 minMultipleJoyce et al. (2004)
    GPMNASA2000.6–2020.12 0.1° × 0.1°30 minMultipleHou et al. (2014)
    GSMaPJAXA2000.3–2020.12 0.1° × 0.1°HourlyMultipleUshio et al. (2009)
    PERSIANNCHRS/UCI2000.3–2020.12 0.25° × 0.25°HourlyInfraredHong et al. (2004)
    FY2PRENSMC/CMA2005.6–2020.12 0.1° × 0.1°HourlyInfraredLu et al. (2004)
    Download: Download as CSV

    Table 3.  Verification metrics

    MetricFull nameUnitOptimal value
    CORRPearson correlation coefficient1
    RMSERoot-mean-square errormm0
    SDRStandard deviation ratio1
    BIASRelative bias%0
    PODProbability of detection1
    FARFalse alarm ratio0
    CSICritical success index1
    ETSEquitable threat score1
    Download: Download as CSV

    Table 4.  Cumulative distribution probability (CDF; %) of PDs and TPA for the key thresholds of daily precipitation intensity derived from multiple SPPs. OBS indicates the observations from rain gauges

    Key thresholdOBSTRMMCMORPHGPMGSMaPPERSIANNFY2PRE
    PD≥ 4 mm day−130.828.921.625.921.216.329.6
    ≥ 30 mm day−11.151.490.931.230.410.211.25
    TPA≥ 4 mm day−180.680.175.278.571.658.979.7
    ≥ 30 mm day−113.215.913.114.75.403.6714.2
    Download: Download as CSV

    Table 5.  The evaluations of daily extreme precipitation events (≥ 30 mm day−1) of SPPs along the STR

    TRMMCMORPHGPMGSMaPPERSIANNFY2PRE
    POD0.2840.3670.3400.7120.0080.395
    FAR0.5020.3880.4440.1270.1660.415
    CSI0.1990.2250.2290.3140.0010.271
    ETS0.1950.2210.2250.3120.0000.267
    Download: Download as CSV

    Table 6.  The CDF (%) of PHs and TPA for the key thresholds of hourly precipitation intensity derived from multiple SPPs

    Key thresholdOBSCMORPHGPMGSMaPPERSIANNFY2PRE
    PH≥ 1 mm h−129.356.938.714.029.520.6
    ≥ 16 mm h−10.922.821.240.00690.0330.36
    TPA≥ 1 mm h−182.693.386.249.873.286.3
    ≥ 16 mm h−116.920.916.10.0180.4716.1
    Download: Download as CSV

    Table 7.  Evaluations of SPPs for the hourly extreme precipitation (16 mm h−1) along the STR

    CMORPHGPMGSMaPPERSIANNFY2PRE
    POD0.0740.1220.5860.1880.142
    FAR0.7490.6760.0020.0670.405
    CSI0.0600.0940.0030.0160.067
    ETS0.0590.0930.0030.0150.066
    Download: Download as CSV

    Table 8.  Relative ranking of SPPs at various timescales. PEDaily represents the daily precipitation events and PEHourly indicates the hourly precipitation events

    DatasetMonthlyDailyHourlyPEDailyPEHourly
    GPM12222
    TRMM345
    CMORPH43444
    GSMaP21313
    PERSIANN66565
    FY2PRE55131
    Download: Download as CSV
  • [1]

    Ahmed, E., F. Al Janabi, J. Zhang, et al., 2020: Hydrologic assessment of TRMM and GPM-based precipitation products in transboundary river catchment (Chenab River, Pakistan). Water, 12, 1902. doi: 10.3390/w12071902.
    [2]

    Amjad, M., M. T. Yilmaz, I. Yucel, et al., 2020: Performance evaluation of satellite- and model-based precipitation products over varying climate and complex topography. J. Hydrol., 584, 124707. doi: 10.1016/j.jhydrol.2020.124707.
    [3]

    Beria, H., T. Nanda, D. S. Bisht, et al., 2017: Does the GPM mission improve the systematic error component in satellite rainfall estimates over TRMM? An evaluation at a pan-India scale. Hydrol. Earth Syst. Sci., 21, 6117–6134. doi: 10.5194/hess-21-6117-2017.
    [4]

    Bhatta, B., S. Shrestha, P. K. Shrestha, et al., 2019: Evaluation and application of a SWAT model to assess the climate change impact on the hydrology of the Himalayan River Basin. CATENA, 181, 104082. doi: 10.1016/j.catena.2019.104082.
    [5]

    Broucke, S. V., H. Wouters, M. Demuzere, et al., 2019: The influence of convection-permitting regional climate modeling on future projections of extreme precipitation: dependency on topography and timescale. Climate Dyn., 52, 5303–5324. doi: 10.1007/s00382-018-4454-2.
    [6]

    Chen, C., Z. Li, Y. N. Song, et al., 2020: Performance of multiple satellite precipitation estimates over a typical arid mountainous area of China: Spatiotemporal patterns and extremes. J. Hydrometeor., 21, 533–550. doi: 10.1175/JHM-D-19-0167.1.
    [7]

    Chen, Z. H., and K. E. Haynes, 2017: Impact of high-speed rail on regional economic disparity in China. J. Trans. Geogr., 65, 80–91. doi: 10.1016/j.jtrangeo.2017.08.003.
    [8]

    de Sousa Afonso, J. M., D. A. Vila, M. A. Gan, et al., 2020: Precipitation diurnal cycle assessment of satellite-based estimates over Brazil. Remote Sens., 12, 2339. doi: 10.3390/rs12142339.
    [9]

    Decker, M., M. A. Brunke, Z. Wang, et al., 2012: Evaluation of the reanalysis products from GSFC, NCEP, and ECMWF using flux tower observations. J. Climate, 25, 1916–1944. doi: 10.1175/JCLI-D-11-00004.1.
    [10]

    Delrieu, G., S. Serrar, E. Guardo, et al., 1999: Rain measurement in hilly terrain with X-band weather radar systems: Accuracy of path-integrated attenuation estimates derived from mountain returns. J. Atmos. Oceanic Technol., 16, 405–416. doi: 10.1175/1520-0426(1999)016<0405:RMIHTW>2.0.CO;2.
    [11]

    Derin, Y., E. Anagnostou, A. Berne, et al., 2019: Evaluation of GPM-era global satellite precipitation products over multiple complex terrain regions. Remote Sens., 11, 2936. doi: 10.3390/rs11242936.
    [12]

    Fu, Y. F., X. Pan, Y. J. Yang, et al., 2017: Climatological characteristics of summer precipitation over East Asia measured by TRMM PR: A review. J. Meteor. Res., 31, 142–159. doi: 10.1007/s13351-017-6156-9.
    [13]

    Gebere, S. B., T. Alamirew, B. J. Merkel, et al., 2015: Performance of high resolution satellite rainfall products over data scarce parts of eastern Ethiopia. Remote Sens., 7, 11,639–11,663. doi: 10.3390/rs70911639.
    [14]

    Guo, J. P., P. M. Zhai, L. Wu, et al., 2014: Diurnal variation and the influential factors of precipitation from surface and satellite measurements in Tibet. Int. J. Climatol., 34, 2940–2956. doi: 10.1002/joc.3886.
    [15]

    Guo, Q. K., S. Y. Cheng, W. Qin, et al., 2020: Vertical variation and temporal trends of extreme precipitation indices in a complex topographical watershed in the Hengduan Mountain Region, China. Int. J. Climatol., 40, 3250–3267. doi: 10.1002/joc.6395.
    [16]

    Gupta, V., M. K. Jain, P. K. Singh, et al., 2020: An assessment of global satellite-based precipitation datasets in capturing precipitation extremes: A comparison with observed precipitation dataset in India. Int. J. Climatol., 40, 3667–3688. doi: 10.1002/joc.6419.
    [17]

    He, Z. H., L. Yang, F. Q. Tian, et al., 2017: Intercomparisons of rainfall estimates from TRMM and GPM multisatellite products over the upper Mekong River basin. J. Hydrometeor., 18, 413–430. doi: 10.1175/JHM-D-16-0198.1.
    [18]

    Hong, Y., K.-L. Hsu, S. Sorooshian, et al., 2004: Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system. J. Appl. Meteor., 43, 1834–1853. doi: 10.1175/JAM2173.1.
    [19]

    Hou, A. Y., R. K. Kakar, S. Neeck, et al., 2014: The Global Precipitation Measurement mission. Bull. Amer. Meteor. Soc., 95, 701–722. doi: 10.1175/BAMS-D-13-00164.1.
    [20]

    Huffman, G. J., 2020: NASA Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG). Algorithm Theoretical Basis Document (ATBD) Version 06, NASA, Greenbelt, MD, 35 pp. Available online at https://gpm.nasa.gov/resources/documents/algorithm-information/IMERG-V06-ATBD. Accessed on 27 September 2022.
    [21]

    Huffman, G. J., R. F. Adler, D. T. Bolvin, et al., 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55. doi: 10.1175/JHM560.1.
    [22]

    Jiang, S. H., R. L. Liu, L. L. Ren, et al., 2020: Evaluation and hydrological application of CMADS reanalysis precipitation data against four satellite precipitation products in the upper Huaihe River basin, China. J. Meteor. Res., 34, 1096–1113. doi: 10.1007/s13351-020-0026-6.
    [23]

    Joyce, R. J., J. E. Janowiak, P. A. Arkin, et al., 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487–503. doi: 10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.
    [24]

    Katiraie-Boroujerdy, P.-S., N. Nasrollahi, K.-L. Hsu, et al., 2013: Evaluation of satellite-based precipitation estimation over Iran. J. Arid Environ., 97, 205–219. doi: 10.1016/j.jaridenv.2013.05.013.
    [25]

    Li, D., K. Yang, W. J. Tang, et al., 2020: Characterizing precipitation in high altitudes of the western Tibetan plateau with a focus on major glacier areas. Int. J. Climatol., 40, 5114–5127. doi: 10.1002/joc.6509.
    [26]

    Li, Z. W., H. Y. Lu, P. Gao, et al., 2020: Characterizing braided rivers in two nested watersheds in the Source Region of the Yangtze River on the Qinghai-Tibet Plateau. Geomorphology, 351, 106945. doi: 10.1016/j.geomorph.2019.106945.
    [27]

    Lin, Z. Q., 2015: Analysis of Tibetan Plateau vortex activities using ERA-Interim data for the period 1979–2013. J. Meteor. Res., 29, 720–734. doi: 10.1007/s13351-015-4273-x.
    [28]

    Lin, Z. Q., W. D. Guo, X. P. Yao, et al., 2021: Tibetan Plateau vortex-associated precipitation and its link with the Tibetan Plateau heating anomaly. Int. J. Climatol., 41, 6300–6313, doi: 10.1002/joc.7195.
    [29]

    Liu, J. Y., J. Du, Y. M. Yang, et al., 2020: Evaluating extreme precipitation estimations based on the GPM IMERG products over the Yangtze River Basin, China. Geomatics Nat. Hazards Risk, 11, 601–618. doi: 10.1080/19475705.2020.1734103.
    [30]

    Liu, K., M. Wang, and T. J. Zhou, 2021: Increasing costs to Chinese railway infrastructure by extreme precipitation in a warmer world. Trans. Res. D, 93, 102797. doi: 10.1016/j.trd.2021.102797.
    [31]

    Liu, S., K. H. Hu, S. J. Zhang, et al., 2019: Comprehensive evaluation of satellite-based precipitation at sub-daily time scales over a high-profile watershed with complex terrain. Earth Space Sci., 6, 2347–2361. doi: 10.1029/2019EA000855.
    [32]

    Lu, C. F., and C. X. Cai, 2019: Challenges and countermeasures for construction safety during the Sichuan–Tibet railway project. Engineering, 5, 833–838. doi: 10.1016/j.eng.2019.06.007.
    [33]

    Lu, D. K., and B. Yong, 2018: Evaluation and hydrological utility of the latest GPM IMERG V5 and GSMaP V7 precipitation products over the Tibetan Plateau. Remote Sens., 10, 2022. doi: 10.3390/rs10122022.
    [34]

    Lu, H. Y., L. D. Ding, Z. Q. Ma, et al., 2020: Spatiotemporal assessments on the satellite-based precipitation products from Fengyun and GPM over the Yunnan-Kweichow Plateau, China. Earth Space Sci., 7, e2019EA000857. doi: 10.1029/2019EA000857.
    [35]

    Lu, N. M., R. You, and W. J. Zhang, 2004: A fusing technique with satellite precipitation estimate and raingauge data. Acta Meteor. Sinica, 18, 141–146.
    [36]

    Lundquist, J., M. Hughes, E. Gutmann, et al., 2019: Our skill in modeling mountain rain and snow is bypassing the skill of our observational networks. Bull. Amer. Meteor. Soc., 100, 2473–2490. doi: 10.1175/BAMS-D-19-0001.1.
    [37]

    Manz, B., S. Páez-Bimos, N. Horna, et al., 2017: Comparative ground validation of IMERG and TMPA at variable spatiotemporal scales in the tropical Andes. J. Hydrometeor., 18, 2469–2489. doi: 10.1175/JHM-D-16-0277.1.
    [38]

    Mishra, A., and M. Rafiq, 2017: Towards combining GPM and MFG observations to monitor near real time heavy precipitation at fine scale over India and nearby oceanic regions. Dyn. Atmos. Oceans, 80, 62–74. doi: 10.1016/j.dynatmoce.2017.10.001.
    [39]

    Navarro, A., E. García-Ortega, A. Merino, et al., 2020: Extreme events of precipitation over complex terrain derived from satellite data for climate applications: An evaluation of the southern slopes of the Pyrenees. Remote Sens., 12, 2171. doi: 10.3390/rs12132171.
    [40]

    Nkunzimana, A., S. B. Bi, M. A. A. Alriah, et al., 2020: Comparative analysis of the performance of satellite-based rainfall products over various topographical unities in central East Africa: Case of Burundi. Earth Space Sci., 7, e2019EA000834. doi: 10.1029/2019EA000834.
    [41]

    Okada, K., and T. Sugiyama, 1994: A risk estimation method of railway embankment collapse due to heavy rainfall. Struct. Safety, 14, 131–150. doi: 10.1016/0167-4730(94)90010-8.
    [42]

    Rahman, K. U., S. H. Shang, M. Shahid, et al., 2018: Developing an ensemble precipitation algorithm from satellite products and its topographical and seasonal evaluations over Pakistan. Remote Sens., 10, 1835. doi: 10.3390/rs10111835.
    [43]

    Sharifi, E., R. Steinacker, and B. Saghafian, 2018: Multi time-scale evaluation of high-resolution satellite-based precipitation products over northeast of Austria. Atmos. Res., 206, 46–63. doi: 10.1016/j.atmosres.2018.02.020.
    [44]

    Sharma, S., Y. Y. Chen, X. Zhou, et al., 2020: Evaluation of GPM-Era satellite precipitation products on the southern slopes of the central Himalayas against rain gauge data. Remote Sens., 12, 1836. doi: 10.3390/rs12111836.
    [45]

    Song, Z. Y., and J. Y. Zhang, 2020: Diurnal variations of summer precipitation linking to the topographical conditions over the Beijing-Tianjin-Hebei region. Sci. Rep., 10, 9701. doi: 10.1038/s41598-020-65743-1.
    [46]

    Soomro, A. G., M. M. Babar, A. Ashraf, et al., 2019: The relationship between precipitation and elevation of the watershed in the Khirthar National Range. Mehran Univ. Res. J. Eng. Technol., 38, 1067–1076. doi: 10.22581/muet1982.1904.18.
    [47]

    Su, J. B., H. S. Lü, D. Ryu, et al., 2019: The assessment and comparison of TMPA and IMERG products over the major basins of Mainland China. Earth Space Sci., 6, 2461–2479. doi: 10.1029/2019EA000977.
    [48]

    Tan, J., G. J. Huffman, D. T. Bolvin, et al., 2019: Diurnal cycle of IMERG V06 precipitation. Geophys. Res. Lett., 46, 13,584–13,592. doi: 10.1029/2019GL085395.
    [49]

    Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res. Atmos., 106, 7183–7192. doi: 10.1029/2000JD900719.
    [50]

    Tong, K., F. G. Su, D. Q. Yang, et al., 2014: Evaluation of satellite precipitation retrievals and their potential utilities in hydrologic modeling over the Tibetan Plateau. J. Hydrol., 519, 423–437. doi: 10.1016/j.jhydrol.2014.07.044.
    [51]

    Ushio, T., K. Sasashige, T. Kubota, et al., 2009: A Kalman filter approach to the Global Satellite Mapping of Precipitation (GSMaP) from combined passive microwave and infrared radiometric data. J. Meteor. Soc. Japan, 87A, 137–151. doi: 10.2151/jmsj.87A.137.
    [52]

    Wang, S. H., J. T. Liu, J. R. Wang, et al., 2019: Evaluation of GPM IMERG V05B and TRMM 3B42V7 precipitation products over high mountainous tributaries in Lhasa with dense rain gauges. Remote Sens., 11, 2080. doi: 10.3390/rs11182080.
    [53]

    Wu, G. X., A. M. Duan, X. Q. Zhang, et al., 2013: Extreme weather and climate changes and its environmental effects over the Tibetan Plateau. Chinese J. Nature, 35, 167–171. (in Chinese)
    [54]

    Xu, J. T., Z. Q. Ma, G. Q. Tang, et al., 2019: Quantitative evaluations and error source analysis of Fengyun-2-based and GPM-based precipitation products over mainland China in summer, 2018. Remote Sens., 11, 2992. doi: 10.3390/rs11242992.
    [55]

    Xue, Y. G., F. M. Kong, S. C. Li, et al., 2021: China starts the world’s hardest “Sky-High Road” project: Challenges and countermeasures for Sichuan-Tibet railway. Innovation, 2, 100105. doi: 10.1016/j.xinn.2021.100105.
    [56]

    Yu, C., D. Y. Hu, M. Q. Liu, et al., 2020: Spatio-temporal accuracy evaluation of three high-resolution satellite precipitation products in China area. Atmos. Res., 241, 104952. doi: 10.1016/j.atmosres.2020.104952.
    [57]

    Yu, L. F., L. X. Ma, H. L. Li, et al., 2020: Assessment of high-resolution satellite rainfall products over a gradually elevating mountainous terrain based on a high-density rain gauge network. Int. J. Remote Sens., 41, 5620–5644. doi: 10.1080/01431161.2020.1734255.
    [58]

    Yu, R. C., J. Li, H. M. Chen, et al., 2014: Progress in studies of the precipitation diurnal variation over contiguous China. J. Meteor. Res., 28, 877–902. doi: 10.1007/s13351-014-3272-7.
    [59]

    Zambrano-Bigiarini, M., A. Nauditt, C. Birkel, et al., 2017: Temporal and spatial evaluation of satellite-based rainfall estimates across the complex topographical and climatic gradients of Chile. Hydrol. Earth Syst. Sci., 21, 1295–1320. doi: 10.5194/hess-21-1295-2017.
    [60]

    Zeng, Q. L., Y. Q. Wang, L. F. Chen, et al., 2018: Inter-comparison and evaluation of remote sensing precipitation products over China from 2005 to 2013. Remote Sens., 10, 168. doi: 10.3390/rs10020168.
    [61]

    Zhang, C. J., H. Y. Wang, J. Zeng, et al., 2020: Short-term dynamic radar quantitative precipitation estimation based on wavelet transform and support vector machine. J. Meteor. Res., 34, 413–426. doi: 10.1007/s13351-020-9036-7.
    [62]

    Zhang, J. P., T. B. Zhao, L. B. Zhou, et al., 2021: Historical changes and future projections of extreme temperature and precipitation along the Sichuan–Tibet Railway. J. Meteor. Res., 35, 402–415. doi: 10.1007/s13351-021-0175-2.
    [63]

    Zhang, S. J., D. H. Wang, Z. K. Qin, et al., 2018: Assessment of the GPM and TRMM precipitation products using the rain gauge network over the Tibetan Plateau. J. Meteor. Res., 32, 324–336. doi: 10.1007/s13351-018-7067-0.
    [64]

    Zhao, T. B., and A. Yatagai, 2014. Evaluation of TRMM 3B42 product using a new gauge-based analysis of daily precipitation over China. Int. J. Climatol., 34, 2749–2762, doi: 10.1002/joc.3872.
    [65]

    Zhou, Y. P., K. Nelson, K. I. Mohr, et al., 2019: A spatial-temporal extreme precipitation database from GPM IMERG. J. Geophys. Res. Atmos., 124, 10,344–10,363. doi: 10.1029/2019JD030449.
  • Xiuping YAO and Zhiqiang LIN.pdf

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Refined Evaluation of Satellite Precipitation Products against Rain Gauge Observations along the Sichuan–Tibet Railway

    Corresponding author: Xiuping YAO, yaoxp@cma.gov.cn
  • 1. Plateau Atmosphere and Environment Key Laboratory of Sichuan Province, School of Atmospheric Sciences, Chengdu University of Information Technology, Chengdu 610225
  • 2. China Meteorological Administration Training Centre, China Meteorological Administration, Beijing 100081
  • 3. Tibet Institute of Plateau Atmospheric and Environmental Science, Lhasa 850000
Funds: Supported by the National Natural Science Foundation of China (42030611 and 42165005), the Second Tibetan Plateau Scientific Expedition and Research (STEP) Program (2019QZKK0103 and 2019QZKK0106), and Key Research and Development Plans of Tibet Autonomous Region in 2022 (XZ202201ZY0008G)

Abstract: Being constructed in southwestern China, the Sichuan–Tibet Railway (STR) travels across the eastern Tibetan Plateau where there is the most complex terrain and changeable weather in the world. Due to sparse ground-based observations over the Tibetan Plateau, precipitation products retrieved by remote sensing are more widely used; however, satellite-based precipitation products (SPPs) have not yet been strictly and systematically evaluated along the STR. This study aims to evaluate the performance of six SPPs by a series of metrics with available ground observations along the STR during 1998–2020. The six SPPs include the datasets derived from the Tropical Rainfall Measuring Mission (TRMM), Climate Prediction Center morphing technique (CMORPH), Global Precipitation Measurement (GPM), Global Satellite Mapping of Precipitation (GSMaP), Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN), and Fengyun-2 satellites precipitation estimate (FY2PRE). The results indicate that most of the SPPs can capture the precipitation characteristics on multiple timescales (monthly, daily, hourly, and diurnal cycle) as shown by the evaluated metrics. The probability density functions of the daily and hourly precipitation are also well represented by the SPPs, and 30 mm day−1 and 16 mm h−1 are identified as the daily and hourly thresholds of extreme precipitation events along the STR. The best SPP varies at different timescales: GPM and GSMaP are suitable for the monthly and daily scale, and FY2PRE and GPM are suited to the hourly scale. In general, GPM is relatively optimum on multiple timescales, and PERSIANN gives the worst performance. In addition, the SPPs perform worse at higher altitudes and for more intense precipitation. Overall, the results from this study are expected to provide essential reference for using the SPPs in meteorological services and disaster prevention in the STR construction and its future operation.

卫星降水产品在川藏铁路区域的精细化评估

川藏铁路将穿越青藏高原东部,这是世界上地形最复杂的地区。复杂多变的降水及相关衍生地质灾害是川藏铁路建设和运营的主要挑战之一。由于当地的地面观测资料稀少,遥感降水产品可以提供至关重要的参考;然而,卫星降水产品(SPP)在川藏铁路沿线的准确性还没有得到很好的评估,这是在川藏铁路沿线使用SPP的瓶颈。本研究旨在评估六种广泛使用的SPP(包括TRMM、CMORPH、GPM、GSMaP、PERSIANN和FY2PRE)的性能;基于1998–2020年的川藏铁路沿线地面降水观测,通过一系列的定量指标给出了六种SPP在多个时间尺度上的表现和相对性能排名。结果表明,大部分SPP可以很好地捕捉到川藏铁路沿线降水的多时间尺度(月、日和小时)的特征和降水的日变化特征。不同时间尺度上表现最佳的SPP有所不同;相对而言,GPM在多个时间尺度的表现是最优的。此外,SPP在高海拔地区和降水强度较大的情况下表现较差。本文的评估结果为SPP应用于川藏铁路建设运营和灾害预防以及提高SPP在复杂地形地区的准确性,提供了重要的参考。
    • The Sichuan–Tibet Railway (STR) is designed to start from Chengdu, pass through Ya’an and Kangding, then enter Tibet via Qamdo, and end in Lhasa. The STR will help develop green tourism, promote local industries and poverty alleviation by increasing the income of local people, and instill a greater sense of national identity among ethnic groups (Chen and Haynes, 2017). The total length of STR is about 1800 km, and it travels through the eastern part of the Tibetan Plateau, where the intricate geological conditions make the construction and operation of STR extremely difficult. Therefore, the STR is one of the most challenging railway projects in the world (Lu and Cai, 2019; Xue et al., 2021). In addition to high altitudes and complex terrain, extreme weather, meteorological disasters, and derived geological hazards (Wu et al., 2013; Lin, 2015; Lin et al., 2021; Liu et al., 2021; Zhang et al., 2021) will seriously threaten the construction and operation of STR. Among meteorological disasters, heavy precipitation can aggravate mountain disasters such as ice avalanches, landslides, and debris flows (Li Z. W. et al., 2020). Thus, accurately monitoring and forecasting of heavy precipitation events at multiple timescales is an essential meteorological service for the STR project.

      During the railway construction, rainwater can erode building materials, such as steel and cement, and reduce the concrete strength, causing the railroad subsidence (Okada and Sugiyama, 1994; Lu and Cai, 2019; Lundquist et al., 2019). Also, it can lead to flooding in tunnels, specifically for the STR with thousands of tunnels to be constructed. The most serious disaster caused by persistent precipitation and heavy precipitation events is geological hazards, which are particularly prone to occur under complex topographic conditions along the STR (Lu and Cai, 2019). Moreover, the precipitation variability is closely related to the local agriculture and animal husbandry, and affects the water supply for millions of Asians living in the downstream areas since many rivers originate or flow through this region, such as the Yarlung Zangbo River, the Nujiang River, the Jinsha River, and the Lantsang River. Hence, accurately monitoring of precipitation in this region is a principal mission of meteorological services.

      Generally, precipitation data can be observed in three ways, i.e., in situ rain gauges, radar networks, and satellite-based precipitation estimations. Due to the complex terrain in the eastern Tibetan Plateau (ETP), there is a great spatial diversity in the precipitation along the STR (Zhang et al., 2018; Zhang et al., 2021). Although the ground-based weather radar can be another choice to monitor precipitation, there are few radars in this area, and the radar detection efficiency and detection range in this complex terrain area are greatly reduced (Delrieu et al., 1999; Zhang et al., 2020). Overall, the sparse ground-based observations cannot meet the requirements of STR construction and operation services. Hence, the satellite-based precipitation products (SPPs) become the most effective choice for monitoring precipitation along the STR.

      The SPPs can be used to capture the extreme precipitation events (Zhou et al., 2019; Gupta et al., 2020; Liu et al., 2020; Navarro et al., 2020), even in complex terrain areas such as the Tibetan Plateau (Tong et al., 2014; Broucke et al., 2019; Chen et al., 2020; Guo et al., 2020; Liu et al., 2020). It suggests the potential availability of the SPPs in capturing the catastrophic rainfall which could seriously jeopardize the STR construction and operation. Besides, the SPPs can supply the gridded regional average precipitation as numerical models do, while in situ rain gauges just reflect local information, particularly in mountainous regions. Therefore, the SPPs may provide a better referential benchmark for regional weather forecast models to improve weather forecasts along the STR.

      The SPPs have been widely used in various regions by meteorologists and hydrologists (Fu et al., 2017; He et al., 2017; Tan et al., 2019; Wang et al., 2019; Ahmed et al., 2020; Jiang et al., 2020). To utilize the SPPs in a specific region, their accuracy should be firstly validated. For this reason, many researchers had conducted numerous evaluations of various SPPs in different regions (Katiraie-Boroujerdy et al., 2013; Gebere et al., 2015; Beria et al., 2017; Manz et al., 2017; Mishra and Rafiq, 2017; Derin et al., 2019; Su et al., 2019; Xu et al., 2019; Nkunzimana et al., 2020; Sharma et al., 2020). Several studies noted that the SPPs also perform well in the complex terrain regions including the Tibetan Plateau (Lu and Yong, 2018; Rahman et al., 2018; Zhang et al., 2018; Bhatta et al., 2019; Liu et al., 2019; Wang et al., 2019; Amjad et al., 2020; Li D. et al., 2020; Li Z. W. et al., 2020; Lu et al., 2020; Yu C. et al., 2020). However, the performance of the SPPs along the STR is yet unclear, especially for the SPPs with multiple timescales. It has been suggested that the performance of the SPPs varies strongly at various timescales (Manz et al., 2017; Sharifi et al., 2018; Zeng et al., 2018; Broucke et al., 2019; Yu C. et al., 2020). Therefore, the evaluation is urgently needed, since it is the basis for the SPPs application along the STR.

      Based on the current deficiencies in evaluating the SPPs along the STR, our study aims to answer the following three questions. First, how do the SPPs, including the average precipitation and the heavy precipitation events, perform in the ETP at multiple timescales? Second, which one performs best at various timescales among the SPPs, to determine which one should be selected for a specific timescale in the weather service for STR construction and operation? Third, how does the complex terrain along the STR affect the performance of the SPPs?

      The remainder of this paper is organized as follows. Section 2 describes the six widely used SPPs, observations from the in situ rain gauges, and the evaluation metrics. Section 3 gives the evaluation results of the SPPs and displays the relative ranks of the SPPs at various timescales along the STR. Section 4 discusses the influence of elevation and precipitation intensity on the performance of SPPs. Finally, the key findings of our study are summarized in Section 5.

    2.   Data and methods
    • As shown in Fig. 1, the ETP can be divided into four sub-regions, i.e., the Sichuan basin (SCB), the West Sichuan (WSC), the eastern Tibet (ET), and the central Tibet (CT). The SCB is located on the east side of the Tibetan Plateau, with an average altitude of less than 1000 m, and it is the first stage of the STR construction (from Chengdu to Ya’an), with the most densely populated region in the study area. The WSC is located in the western part of Sichuan Province and is the eastward extension of the Tibetan Plateau, with an average altitude of about 3400 m. The section of the STR in the WSC is from Ya’an to Litang. The ET mainly includes Nyingchi and Qamdo in the eastern part of the Tibet Autonomous Region, with an average altitude of 4025 m. In the ET, the STR travels through the famous Hengduan Mountains, which have the greatest altitude gradient in the study area. The CT, in the central part of the Tibet Autonomous Region and the Tibetan Plateau, is the destination of the STR. The CT mainly includes Lhasa, Shannan, and Nagqu, with an average altitude of 4348 m.

      Figure 1.  (a) Distributions of the elevation (shading), automatic weather stations (AWSs; black dots), and national basic stations (red circles) over the eastern TP (red rectangles) along the Sichuan–Tibet Railway (STR; thick blue line). (b) The histogram (%) of the elevation of the rain gauges in the entire region along the STR and each sub-region, and the vertical red line denotes the average elevation of the region.

      The elevation of the study area is obtained from the global digital elevation model data, with a resolution of 30 arc s, and is provided by the website of the U.S. Geological Survey (https://lta.cr.usgs.gov/GTOPO30).

    • Figure 1 shows the basic information about the rain gauges along the STR, including the national basic observation (NBO) and the automatic weather station (AWS) networks. The data periods of the NBO and AWS are from 1998 and 2015 to the present, respectively. The density of AWSs is lower than the spatial resolution of most SPPs, even in the SCB where the observation is the most intensive. Especially, the resolution of the NBO observation networks over the entire ETP with long-term records is much coarser than that of SPPs (Table 1). More unfortunately, since most of the study area is sparsely populated, a large percentage of AWSs are unable to work stably due to the lack of maintenance and repair. Compared with the sparse ground-based observations along the STR, SPPs can provide long-term, continuous, and refined information for meteorological services.

      RegionAverage altitude (m)Altitude range (m)AWS/NBO countAverage spatial density of rain gauges for
      AWS/NBO networks (No. per 100 km2)
      ETP2131 284–55802678/610.303/0.0070
      SCB 985 284–28511551/110.785/0.0072
      WSC3394 515–4536 695/220.289/0.0088
      ET40251618–5193 197/150.105/0.0075
      CT43483294–5580 235/130.120/0.0085

      Table 1.  Basic information about the ground-based rain gauges in the whole eastern Tibetan Plateau (ETP) along the STR, and its each sub-region. SCB indicates the Sichuan basin, WSC the western part of Sichuan, ET the eastern part of Tibet, CT the central part of Tibet, AWS the automatic weather station, and NBO the national basic observation

      The evaluations of the daily and monthly precipitation and daily heavy precipitation events are conducted by using the data from the NBO network during 1998–2020. Since the AWS network has an observation period and cannot detect snowfall, the evaluations of the hourly precipitation and hourly heavy precipitation events are performed by the data from AWSs in the warm seasons (May–September) of 2015–2020. The daily precipitation is defined as the accumulative value from 0000 to 2359 UTC, and the hourly precipitation is cumulative from NN00 to NN59 UTC in the hour NN, and thus the ground-based precipitation and SPPs can be compared within the same period. The rain gauge data are provided by the National Meteorological Information Centre, China Meteorological Administration (CMA), and can be obtained from the China Integrated Meteorological Information Service System.

    • (1) TRMM

      The Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis data are used in this research. TRMM was a research satellite that operated during 1997–2015 to help study the distribution and variability of precipitation within the tropics as a part of the water cycle in the climate system. In coordination with other satellites in NASA’s Earth Observing System, TRMM provided important precipitation information by using several spaceborne instruments, facilitating the study of the interactions among water vapor, clouds, and precipitation that are critical factors in regulating climate. The TRMM datasets can be downloaded from the website https://gpm.nasa.gov/missions/trmm.

      (2) CMORPH

      The NOAA Climate Prediction Center morphing technique (CMORPH) product is a precipitation estimation product. This technique is a flexible method that combines the existing passive microwave-based precipitation estimation from multiple low orbit satellites with the infrared data from multiple geostationary satellites (Joyce et al., 2004). Here, the CMORPH bias-corrected dataset (https://www.cpc.ncep.noaa.gov/products) is used to evaluate its performance along the STR.

      (3) GPM

      The Integrated Multisatellite Retrievals for Global Precipitation Measurement (GPM) Mission (IMERG) version 06 (https://pmm.nasa.gov/GPM) is the next generation product of TRMM. The IMERG fuses the early precipitation estimations collected by the TRMM satellite (2000–2015) with the more recent precipitation estimations collected by the GPM satellite (2014–present). The version of IMERG Final Run is evaluated in this study at the monthly, daily, and hourly scales (Huffman, 2020).

      (4) GSMaP

      The Global Satellite Mapping of Precipitation (GSMaP) product (http://sharaku.eorc.jaxa.jp/GSMaP) is produced by the Japan Aerospace Exploration Agency (JAXA), which aims to develop an accurate and high-resolution global precipitation product based on the passive microwave and infrared satellites. The data of GSMaP-Gauge version 6 are evaluated in this study.

      (5) PERSIANN

      The Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN; http://chrsdata.eng.uci.edu) is developed by the Center for Hydrometeorology and Remote Sensing (CHRS) at the University of California, Irvine (UCI). The PERSIANN is the estimation of rainfall rate computed by the neural network function classification/approximation procedures and based on the infrared brightness temperature image from geostationary satellites.

      (6) FY2PRE

      The Precipitation Estimation from Fengyun-2 (FY2PRE) series geostationary satellites (FY-2C–FY-2H) is produced by the National Satellite Meteorological Centre (NSMC) of CMA and can be obtained from http://satellite.nsmc.org.cn. The FY2PRE covers the period from June 2005 to December 2020. The SPPs derived from the newer satellite will be used in days when multiple satellites overlap. The monthly precipitation data of the FY2PRE are accumulated by the daily precipitation.

      The GPM, GSMaP, and FY2PRE products have the resolution of 0.1° × 0.1° (approximately equivalent to the density of 1 per 100 km2), and the resolutions of TRMM, CMORPH, and PERSIANN products are 0.25° × 0.25° (approximately equivalent to the spatial density of 0.16 per 100 km2). The information about the SPPs is presented in Table 2.

      DatasetInstitutionPeriodResolutionScaleSensorReference
      TRMMNASA1998.1–2019.12 0.25° × 0.25°3 hMultipleHuffman et al. (2007)
      CMORPHCPC/NOAA1998.1–2020.120.25° × 0.25°30 minMultipleJoyce et al. (2004)
      GPMNASA2000.6–2020.12 0.1° × 0.1°30 minMultipleHou et al. (2014)
      GSMaPJAXA2000.3–2020.12 0.1° × 0.1°HourlyMultipleUshio et al. (2009)
      PERSIANNCHRS/UCI2000.3–2020.12 0.25° × 0.25°HourlyInfraredHong et al. (2004)
      FY2PRENSMC/CMA2005.6–2020.12 0.1° × 0.1°HourlyInfraredLu et al. (2004)

      Table 2.  Satellite-based precipitation products. The full names of the institutes and datasets are detailed in Section 2.3

    • To quantify the performance of various products, we adopted several statistical metrics in this research (Table 3). The first four metrics, including the Pearson correlation coefficient (CORR), the relative bias (BIAS), the root-mean-square error (RMSE), and the standard deviation ratio (SDR), are calculated as follows [Eqs. (1)–(4)].

      MetricFull nameUnitOptimal value
      CORRPearson correlation coefficient1
      RMSERoot-mean-square errormm0
      SDRStandard deviation ratio1
      BIASRelative bias%0
      PODProbability of detection1
      FARFalse alarm ratio0
      CSICritical success index1
      ETSEquitable threat score1

      Table 3.  Verification metrics

      $$ \mathrm{C}\mathrm{O}\mathrm{R}\mathrm{R}=\frac{\sum _{i=1}^{N}\left({{y}}_{i}-\bar{{y}}\right)\left({{x}}_{i}-\bar{{x}}\right)}{\sqrt{\sum _{i=1}^{N}{\left({{y}}_{i}-\bar{{y}}\right)}^{2}} {\sqrt {\sum _{i=1}^{N}{\left({{x}}_{i}-\bar{{x}}\right)}^{2}}}}, $$ (1)
      $$ \mathrm{B}\mathrm{I}\mathrm{A}\mathrm{S}=\frac{\sum _{i=1}^{N}\left({{y}}_{i}-{{x}}_{i}\right)}{\sum _{i=1}^{N}{{x}}_{i}}\times 100\mathrm{\%}, $$ (2)
      $$ \rm{RMSE}=\sqrt{\frac{\sum _{i=1}^{N}{\left({{y}}_{i}-{{x}}_{i}\right)}^{2}}{N}}, $$ (3)
      $$ \rm{SDR}=\sqrt{\frac{\sum _{i=1}^{N}{\left({{y}}_{i}-\bar{{y}}\right)}^{2}}{\sum _{i=1}^{N}{\left({{x}}_{i}-\bar{{x}}\right)}^{2}}}, $$ (4)

      where N is the total number of samples, and x and y denote the observed samples and the estimated samples, respectively.

      The last four metrics, namely, the probability of detection (POD), the false alarm ratio (FAR), the critical success index (CSI), and the equitable threat score (ETS), are used to evaluate the performance of SPPs for the daily and hourly precipitation events. The calculation equations are as follows [Eqs. (5)–(8)].

      $$ \mathrm{P}\mathrm{O}\mathrm{D}=\frac{{N}_{11}}{{N}_{11}+{N}_{01}}, $$ (5)
      $$ \mathrm{F}\mathrm{A}\mathrm{R}=\frac{{N}_{10}}{{N}_{11}+{N}_{10}}, $$ (6)
      $$ \mathrm{C}\mathrm{S}\mathrm{I}=\frac{{N}_{11}}{{N}_{11}+{N}_{10}+{N}_{01}}, $$ (7)
      $$\hspace{35pt} \mathrm{E}\mathrm{T}\mathrm{S}=\frac{{N}_{11}-\mathrm{R}\mathrm{N}\mathrm{D}}{{N}_{11}+{N}_{10}+{N}_{01}-\mathrm{R}\mathrm{N}\mathrm{D}}, $$ (8)

      where N is the number of samples, N11 the precipitation from gauges and satellites, N10 the precipitation only from satellites, N01 the precipitation only from gauges, and N00 the precipitation not observed by both gauges and satellites. POD is used to measure the precipitation events correctly detected by satellites, while FAR is applied to measure the precipitation events incorrectly detected. CSI and ETS are used to measure the correspondence between SPPs and in situ gauges, and ETS is modified by an adjustment coefficient of RND [Eq. (9)].

      $$ \mathrm{R}\mathrm{N}\mathrm{D}=\frac{{(N}_{11}+{N}_{10}\left)\right({N}_{11}+{N}_{01})}{N}. $$ (9)

      Due to the sparse observations and the uneven precipitation distribution caused by the complex terrain along the STR, the SPP data are interpolated into stations through the bilinear interpolation so as to obtain the verification metrics. In addition, we have tested the method of the nearest grid cell to the stations, and find that there is no significant difference to the bilinear method. An evaluation method proposed by Zhao and Yatagai (2014) has also been tested, and it also presents similar results for the performance of SPPs.

    • The objective ranking system proposed by Decker et al. (2012) is employed to rank the comprehensive performance of each SPP at a specific timescale, and the calculation method is as Eq. (10).

      $$ \mathrm{S}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}\left[\mathrm{S}\mathrm{P}\mathrm{P},\mathrm{S}\mathrm{t}\mathrm{a}\left(i\right),\mathrm{S}\mathrm{t}\mathrm{n}\left(j\right)\right]=100\times \frac{{{M}}_{\mathrm{S}\mathrm{P}\mathrm{P}}-{{M}}_{\mathrm{W}\mathrm{o}\mathrm{r}\mathrm{s}\mathrm{t}}}{{{M}}_{\mathrm{B}\mathrm{e}\mathrm{s}\mathrm{t}}-{{M}}_{\mathrm{W}\mathrm{o}\mathrm{r}\mathrm{s}\mathrm{t}}}, $$ (10)

      where Stn(j) represents the jth station (j = 1, ..., N), and Sta(i) denotes the score of the ith statistical metric (i = 1, ..., 4). The rankings of the yearly, monthly, daily, and diurnal-cycle assessments are performed by CORR, RMSE, SDR, and BIAS, and the rankings of the heavy precipitation events are conducted by POD, FAR, CSI, and ETS. Here, ${{M}}_{\mathrm{B}\mathrm{e}\mathrm{s}\mathrm{t}}$ and ${{M}}_{\mathrm{W}\mathrm{o}\mathrm{r}\mathrm{s}\mathrm{t}}$ respectively indicate the SPPs with the best and worst performance at the jth station measured by the ith metric; ${{M}}_{\mathrm{S}\mathrm{P}\mathrm{P}}$ is the metric of each SPP. Thus, the SPP with the best (worst) performance receives a score of 100 (0), and the other SPPs receive a score of 0–100.

      $$ {\rm Ranking}\left(\mathrm{S}\mathrm{P}\mathrm{P}\right) = {\sum }_{i=1}^{4}{\sum }_{j=1}^{N}\mathrm{S}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}[\mathrm{S}\mathrm{P}\mathrm{P},\mathrm{S}\mathrm{t}\mathrm{a}\left(i\right),\mathrm{S}\mathrm{t}\mathrm{n}\left(j\right)].\hspace{-3pt} $$ (11)

      The ranking of an SPP is based on the comprehensive score against the gauges.

    3.   Verification results
    • Figure 2 displays the spatial distributions of mean yearly precipitation from rain gauges (Fig. 2a) and SPPs (Figs. 2b–g) along the STR. Most of the SPPs can well characterize the major spatial distribution of precipitation in the study area, except the PERSIANN. The precipitation distribution along the STR shows a decrease from southeast to northwest, with a local minimum over the central STR. In the study area, the microwave-based SPPs (TRMM, GPM, and GSMaP) better capture the spatial pattern of precipitation than the others. The CMORPH overestimates the rain-shadow effect of the Himalayas and Hengduan Mountains, and also has significant precipitation overestimations in the central Tibetan Plateau. The two SPPs (PERSIANN and FY2PRE) based on infrared data have a significant underestimation on the southeast side of the Tibetan Plateau and the SCB. Note that the PERSIANN shows a spurious heavy precipitation center over the northeastern Tibet Autonomous Region and southern Qinghai Province.

      Figure 2.  Spatial variability of annual accumulated precipitation along the STR from (a) in situ rain gauges (the legend is the same as the color bar), (b) TRMM, (c) CMORPH, (d) GPM, (e) GSMaP, (f) PERSIANN, and (g) FY2PRE. The red thick line in each figure denotes the STR.

      Figure 3 shows the scatterplots of monthly SPPs versus the observations along the STR. The microwave-based SPPs (GPM, TRMM, and GSMaP) have almost the optimal correlation coefficients (more than 0.95) with monthly observations, while the infrared-based SPPs (PERSIANN and FY2PRE) have relatively low correlation coefficients (only about 0.6) with monthly observations. The CMORPH, GSMaP, and FY2PRE slightly underestimate the monthly precipitation (about −2% of the bias), while the GPM and TRMM show a slight overestimation. Moreover, the PERSIANN greatly overestimates the monthly precipitation compared with others.

      Figure 3.  Scatterplots of monthly precipitation from the gauge observation along the STR versus satellite-based precipitation products (SPPs) including (a) TRMM, (b) CMORPH, (c) GPM, (d) GSMaP, (e) PERSIANN, and (f) FY2PRE. The dash and solid lines denote best fitted performance (y = x) and linear fitting of the SPPs to observations, respectively.

      Figure 4 presents the Taylor diagrams of the correlation coefficient (i.e., CORR), the centered RMSE, and the SDR between SPPs and rain gauge observations at the yearly and seasonal scales (Taylor, 2001). The CORR, RMSE, and SDR metrics are calculated by yearly and seasonal precipitation from rain gauges and SPPs. The centered RMSE has the same statistical significance as the RMSE, and its value ranges from 0 to 1. Therefore, it is more convenient to use this measurement in a Taylor diagram. In Fig. 4, the closer the distance between the centered value and the observation point (red point), the better the SPP performance is. The GPM, GSMaP, and TRMM have relatively good performance, and the CMORPH exhibits medium performance, but the FY2PRE and PERSIANN perform relatively poorly. There is no obvious pattern in the seasonal differences, with some SPPs performing better in the warm season and some in the cold season. In terms of the overall SPP accuracy at yearly and seasonal scales, the best performance is achieved in the SCB, followed by the WSC, and the worst is in the ET and CT. The performance of SPPs is closely related to the regional altitude. This will be discussed in Section 4, and it can be found that the above characteristics are more pronounced in summer.

      Figure 4.  Taylor diagrams for correlation coefficients (CORR), standard deviation ratios (SDR), and normalized root-mean-square errors (RMSE) on the (a) entire ETP along the STR and (b) SCB, (c) WSC, (d) ET, and (e) CT.

    • The density-color scatterplots and the quantitative comparison between SPPs and gauge observations for the daily precipitation along the STR (Fig. 5) show that all SPPs tend to overestimate daily precipitation in light precipitation days and underestimate that in heavy precipitation days. On average, all SPPs perform pretty well with daily CORR values of more than 0.6, except for the PERSIANN. The TRMM, GPM, and FY2PRE manifest slightly positive BIAS values of less than 10%, the CMORPH and GSMaP get small negative BIAS values of less than −5%, and the PERSIANN largely underestimates the precipitation (BIAS close to −50%). The GSMaP gets the smallest RMSE of less than 4 mm among all SPPs, while the RMSEs of the other five SPPs are about 5 mm. The RMSE is mainly derived from no-precipitation days, i.e., SPPs assign the observed no-precipitation day to a significant rainfall day, and conversely, treat the precipitation days as the no-precipitation days.

      Figure 5.  The two-dimensional distribution density of daily precipitation from the in situ rain gauges along the STR versus the SPPs of (a) TRMM, (b) CMORPH, (c) GPM, (d) GSMaP, (e) PERSIANN, and (f) FY2PRE.

      The probability distribution function (PDF) and cumulative distribution probability (CDF) of precipitation days and total precipitation amount can present useful information about the frequency histogram of a dataset. The PDF and CDF of the precipitation from rain gauges and SPPs along the STR are shown in Fig. 6. All of the SPPs and rain gauge data show that the highest percentage occurs at the rainfall intensity between 0 and 1 mm day−1, accounting for about 40% of the precipitation days. Compared with rain gauge observations, SPPs tend to underestimate the frequency of trace rain days (0.1 mm day−1) and overestimate the frequency of slight rain days (0.2–5 mm day−1). The GSMaP and PERSIANN remarkably overestimate the contribution of light rain (0.1–8 mm day−1) and underestimate the contribution of heavy rain, while the other four SPPs somewhat present the PDF and CDF of the precipitation amount.

      Figure 6.  (a, c) Probability distribution function (PDF) and (b, d) cumulative distribution probability (CDF) of (a, b) daily precipitation days (PDs) and (c, d) total precipitation amount (TPA) with varying intensities over the entire ETP along the STR. The dashed lines denote the dominate precipitation intensities for the TPA (4 mm day−1) and the daily extreme precipitation (30 mm day−1).

      The values of two crucial daily precipitation intensities, i.e., 4 and 30 mm day−1 (Table 4), are detected in Fig. 6, where 4 mm day−1 is the threshold contributing to the dominant precipitation amount. The precipitation days with the intensity of ≥ 4 mm day−1 account for about 20%–30% of the total precipitation days and contribute more than 80% to the total precipitation amount.

      Key thresholdOBSTRMMCMORPHGPMGSMaPPERSIANNFY2PRE
      PD≥ 4 mm day−130.828.921.625.921.216.329.6
      ≥ 30 mm day−11.151.490.931.230.410.211.25
      TPA≥ 4 mm day−180.680.175.278.571.658.979.7
      ≥ 30 mm day−113.215.913.114.75.403.6714.2

      Table 4.  Cumulative distribution probability (CDF; %) of PDs and TPA for the key thresholds of daily precipitation intensity derived from multiple SPPs. OBS indicates the observations from rain gauges

      Figure 7 presents the performance of SPPs for the precipitation days with the intensity of more than 4 mm day−1. Compared with other SPPs, the GSMaP performs better with higher POD, CSI, and ETS, as well as lower FAR. Conversely, the PERSIANN shows the worst performance. The POD, CSI, and ETS of the GPM are similar to those of the TRMM and CMORPH, but the FAR is lower. The POD of the FY2PRE is higher in the CT than that of the TRMM, CMORPH, and GPM. In general, SPPs perform better along the eastern STR (SCB and WSC) than along the western part (CT and ET), which may be due to the altitude difference.

      Figure 7.  Spatial distributions of statistical metrics for the SPP daily precipitation of ≥ 4 mm day−1 along the STR: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The panels from top to bottom show the results of the TRMM, CMORPH, GPM, GSMaP, PERSIANN, and FY2PRE, respectively.

      The days with a precipitation intensity of more than 30 mm day−1 are about 1% of the total precipitation days, and they are considered as the daily extreme precipitation events along the STR. These extreme precipitation events contribute about 15% of the total precipitation amount, except the GSMaP and PERSIANN that prominently underestimate the CDF of precipitation days and total precipitation amount with the precipitation intensity of more than 30 mm day−1. Table 5 shows the performance of SPPs for the daily extreme precipitation events. The performance of GSMaP and PERSIANN is the best and worst among the SPPs, respectively. The performance of TRMM is slightly worse than that of CMORPH. The performance ranking of SPPs is significantly affected by the daily precipitation intensity.

      TRMMCMORPHGPMGSMaPPERSIANNFY2PRE
      POD0.2840.3670.3400.7120.0080.395
      FAR0.5020.3880.4440.1270.1660.415
      CSI0.1990.2250.2290.3140.0010.271
      ETS0.1950.2210.2250.3120.0000.267

      Table 5.  The evaluations of daily extreme precipitation events (≥ 30 mm day−1) of SPPs along the STR

    • The density-color scatterplots of SPPs versus the gauge observations along the STR for the hourly precipitation (Fig. 8) show that all SPPs tend to overestimate the hourly precipitation in light rain events and underestimate it in heavy rain events. The CMORPH and GPM show substantial overestimation (BIAS of more than 100%), particularly in the smaller precipitation ranges. The GSMaP slightly overestimates the hourly precipitation (BIAS of about 10%), while the PERSIANN and FY2PRE have a slight underestimation. The FY2PRE and PERSIANN generally underestimate the hourly precipitation intensity in all the precipitation ranges, while the others tend to overestimate the hourly precipitation in the smaller precipitation ranges and underestimate it in the larger precipitation ranges. The significant overestimation makes the large RMSEs of CMORPH and GPM, more than 2 mm h−1, while the RMSEs of others are less than 1.5 mm h−1. The CORR of the GSMaP is the largest, more than 0.4, followed by the GPM. The FY2PRE has the smallest RMSE and relatively small CORR.

      Figure 8.  The two-dimensional distribution density of hourly precipitation from in situ rain gauges along the STR versus SPPs: (a) CMORPH, (b) GPM, (c) GSMaP, (d) PERSIANN, and (e) FY2PRE.

      Figure 9 shows the PDF and CDF of precipitation hours and the total precipitation amount derived from SPPs and rain gauges. The consistency between SPPs and gauge observations is significantly lower than that of the daily precipitation. SPPs and gauge data show that the highest percentage occurs at the rainfall intensity of less than 1 mm h−1, accounting for about 60%–85% of the total precipitation hours, except for the CMORPH (about 40%). The GSMaP greatly overestimates the frequency of light rain (< 1 mm h−1), which accounts for 85% and 50% of the total precipitation hours and total precipitation amount, respectively, compared with only 65% and 15% for the observations. Conversely, the CMORPH greatly underestimates the frequency of light and moderate rain.

      Figure 9.  As in Fig. 6, but for the hourly precipitation (PHs: precipitation hours). The dash lines denote the precipitation intensities of 1 and 16 mm h−1.

      Table 6 shows the intensity of the dominant precipitation that contributes most of the total precipitation amount (1 mm h−1) and the threshold (16 mm h−1) of hourly extreme precipitation events identified from Fig. 9. The results suggest that in terms of gauge observation, the number of hours with a precipitation intensity of 1 mm h−1 accounts for about 30% of the total precipitation hours, and its precipitation contributes more than 80% of total precipitation amount. The GSMaP remarkably underestimates the percentage of precipitation hours (14%) and the precipitation amount (less than 50%), and the CMORPH highly overestimates them (about 57% for precipitation hours and more than 93% for the precipitation amount). The other three SPPs (GPM, PERSIANN, and FY2PRE) have similar CDF values for the hourly precipitation of more than 1 mm h−1. These three SPPs occupy 20%–39% of the precipitation hours and contribute 73%–86% to the total precipitation.

      Key thresholdOBSCMORPHGPMGSMaPPERSIANNFY2PRE
      PH≥ 1 mm h−129.356.938.714.029.520.6
      ≥ 16 mm h−10.922.821.240.00690.0330.36
      TPA≥ 1 mm h−182.693.386.249.873.286.3
      ≥ 16 mm h−116.920.916.10.0180.4716.1

      Table 6.  The CDF (%) of PHs and TPA for the key thresholds of hourly precipitation intensity derived from multiple SPPs

      Figure 10 presents the performance of SPPs for the hourly precipitation intensity of more than 1 mm h−1 along the STR. It can be found that the CMORPH and the GPM have higher FAR than other SPPs. The POD values of FY2PRE and GSMaP are the highest and the second-highest compared with other SPPs, respectively. The FAR of FY2PRE is larger than that of GSMaP, particularly in the western part of the STR. Overall, the FY2PRE has the best performance for the CSI and ETS metrics, followed by the GSMaP. Same as the daily timescale, SPPs perform better along the eastern STR than the western part (CT and ET).

      Figure 10.  The spatial distributions of statistical metrics for the SPP hourly precipitation of ≥ 1 mm h−1 along the STR: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The panels from top to bottom show the results of CMORPH, GPM, GSMaP, PERSIANN, and FY2PRE, respectively.

      The hours of observed precipitation with the intensity of more than 16 mm h−1 occupy about 1% of the total precipitation hours, and thus the precipitation with such intensity is referred to as hourly extreme precipitation events along the STR (as shown in Table 6). These extreme precipitation events contribute about 16%–20% of precipitation for gauge observations and SPPs, except for the GSMaP and PERSIANN that account for less than 1% of the hourly extreme precipitation events. Table 7 shows the performance of SPPs for the hourly extreme precipitation events, and the performance is different for various verification metrics. Although the GSMaP has the highest POD and the lowest FAR, it performs the worst for the CSI and ETS metrics due to the significant underestimation for the PDF of hourly extreme precipitation events. The FY2PRE performs well for POD and FAR, but has relatively worse CSI and ETS metrics than the GPM. The CSI and ETS of the GPM are the highest, but it performs the worst for POD and FAR.

      CMORPHGPMGSMaPPERSIANNFY2PRE
      POD0.0740.1220.5860.1880.142
      FAR0.7490.6760.0020.0670.405
      CSI0.0600.0940.0030.0160.067
      ETS0.0590.0930.0030.0150.066

      Table 7.  Evaluations of SPPs for the hourly extreme precipitation (16 mm h−1) along the STR

    • As the diurnal variation is the basic characteristic of precipitation, the performance of SPPs on the diurnal cycle is an important factor in measuring their capabilities (Guo et al., 2014; Yu et al., 2014; Tan et al., 2019; de Sousa Afonso et al., 2020; Song and Zhang, 2020). Figure 11 shows the relative ratios derived from SPPs and gauge observations in the sub-regions along the STR at the local solar time (LST). In the SCB, which is a flat plain, SPPs can well reflect the diurnal cycle of precipitation, including the peak in the evening (at about 1900 LST) and the valley at about 0900 LST, and the precipitation occurring at nighttime (1900–0700 LST) contributes about 73% of the total precipitation both for SPPs and gauge observations. In the WSC, the precipitation ratio is the highest at night, more than 80%. SPPs can well capture this feature, with the precipitation ratio of 72%–81% at night, although SPPs exhibit one hour ahead of the peak and valley time compared with the observations. In the ET and CT, the peak time of SPPs is advanced earlier to around 2000 LST compared with the peak time of about 0000 LST for the observations from rain gauges. In addition, SPPs identify one hour ahead of the valley time compared with gauge observations. SPPs overestimate the precipitation contribution in the early morning and underestimate that in the late nighttime in the ET and CT, thus well characterizing the relative ratios of the precipitation in the nighttime in these two regions.

      Figure 11.  The diurnal cycle of precipitation from the AWS network and SPPs in the warm season (May–September) along the STR and in the sub-regions: (a) SCB, (b) WSC, (c) ET, and (d) CT. The gray shaded area denotes the hours in the nighttime.

      Although the various SPPs perform quite differently at the hourly timescale, they exhibit highly similar features in the diurnal cycle. Except for the GSMaP, it appears a particular valley at the late nighttime (about 0400–0600 LST), which is different with other SPPs. Overall, as demonstrated by SPPs and gauge observations, the precipitation along the STR is concentrated at night (more than 70% of total precipitation) in all sub-regions. Compared with the situations at the eastern side of the Tibetan Plateau (SCB and WSC), the performance of SPPs is worse in characterizing the diurnal cycle of precipitation on the Tibetan Plateau (ET and CT), indicating that the complex terrain affects not only the precipitation intensity but also the diurnal cycle reproduced by SPPs (Song and Zhang, 2020).

    • The relative ranking of SPPs at various timescales is presented in Table 8. The GPM performs the best at monthly timescale and the second-best at daily and hourly timescales. The GSMaP has perfect performance at daily timescale. The FY2PRE has the best performance at hourly timescale and for the hourly precipitation events but performs worse in the other evaluations. The previous evaluations on SPPs in the other regions (Xu et al., 2019; Lu et al., 2020) also suggested that the FY2PRE performs better than the GPM at hourly timescale and has worse performance at daily and monthly timescales, indicating that the varying performance of FY2PRE is not a regional feature but an overall one. As a subsequent product of TRMM, the GPM performs better than TRMM and CMORPH among all evaluations. Due to the observational limitations to build a model in the study area, the performance of PERSIANN is worse than other SPPs. Although SPPs perform differently at multiple timescales, the GPM is generally the relative optimal choice since it performs well in all evaluations. Although the daily evaluation metrics show that the GSMaP performs better than GPM, the GSMaP highly underestimates the heavy precipitation days, and thus its performance is worse than the GPM. In addition, the best performance of FY2PRE may be affected by more in situ observations, which is more integrated than the other SPPs.

      DatasetMonthlyDailyHourlyPEDailyPEHourly
      GPM12222
      TRMM345
      CMORPH43444
      GSMaP21313
      PERSIANN66565
      FY2PRE55131

      Table 8.  Relative ranking of SPPs at various timescales. PEDaily represents the daily precipitation events and PEHourly indicates the hourly precipitation events

    4.   Discussion: factors relating to SPP performance
    • Terrain effect may be a crucial factor that influences SPP performance, especially on the Tibetan Plateau (Zambrano-Bigiarini et al., 2017; Zhang et al., 2018; Soomro et al., 2019; Yu L. F. et al., 2020), and the variations in the evaluation metrics at daily and hourly timescales with increasing elevation are investigated in this research. For the daily precipitation evaluation (not shown), the CORR of SPPs decreases with increasing elevation, except for the GSMaP. The PERSIANN highly underestimates the daily precipitation, with a large BIAS, while the GSMaP is nearly to be perfect (BIAS close to 0.0) and varies slightly in pace with the elevation. The BIAS values of the other four SPPs show that they underestimate precipitation at lower altitudes and overestimate at higher altitudes. The RMSE of GSMaP is the lowest among all the SPPs and slightly decreases with increasing elevation, while the RMSE values of the other five SPPs are more significant at higher altitudes. The RMSE of FY2PRE increases the fastest among all SPPs with increasing elevation. The SDR shows that in the regions with higher altitudes, all SPPs overestimate the standard deviation of daily precipitation compared with the regions with lower altitudes.

      The same features are identified from the verified metrics for the major precipitation events (≥ 4.0 mm day−1 and 1 mm h−1; Fig. 12 shows the result of daily precipitation ≥ 4.0 mm day−1). The POD, CSI, and ETS decrease with increasing elevation, and the FAR enlarges in pace with increasing elevation. In contrast, the POD, CSI, and ETS of all SPPs for the hourly precipitation events decrease with the increase of elevation. This result reveals the possibility of improving the accuracy of satellite rainfall retrieval in regions with complex terrain. Overall, since the capabilities of SPPs to detect precipitation events decrease with increasing elevation, and there is a lack of observations, SPPs are more demanding in the higher terrain regions. Therefore, it is more imperative to improve the quality of SPPs in the mountainous regions.

      Figure 12.  The elevation against the evaluation metrics of the daily precipitation: (a) POD, (b) FAR, (c) CSI, and (d) ETS. The dash and solid lines indicate the linear regression of the metric to the elevation, and the dash (solid) line denotes that the metric decreases (increases) with increasing elevation.

    • The SPP performance is strongly affected by the precipitation intensity. The correlation of daily metrics with the precipitation intensity is presented in Fig. 13. The POD, CSI, and ETS of GSMaP are the largest, while those of PERSIANN are the least. Generally, the POD, CSI, and ETS of SPPs decrease with increasing precipitation intensity, and remain constant after the intensity reaching about 25 mm day−1. The FAR of GSMaP is the lowest among all SPPs and changes insignificantly with increasing precipitation intensity. The FAR of TRMM, CMORPH, GPM, and FY2PRE increases with increasing precipitation intensity and remains constant after the intensity reaching 25 mm day−1, while the FAR of PERSIANN decreases with the increase of precipitation intensity.

      Figure 13.  Scores of the (a) POD, (b) FAR, (c) CSI, and (d) ETS for SPPs with various daily precipitation intensities.

      The relationship between the hourly precipitation intensity and the evaluation metrics demonstrates the similar feature as the daily precipitation (not shown). The POD, CSI, and ETS of SPPs decrease with the increase of hourly precipitation intensity. The FAR of PERSIANN and GSMaP increases with the increasing hourly precipitation intensity, and that of the other three SPPs exhibits the opposite curve. It is caused by the significant underestimation of the hourly heavy precipitation in the PERSIANN and GSMaP. The relationships between verification metrics and the hourly precipitation intensity suggest that the ability of SPPs to detect heavy precipitation decreases.

    5.   Conclusions
    • To determine the SPP performance along the STR, we select six widely used products, i.e., TRMM, CMORPH, GPM, GSMaP, PERSIANN, and FY2PRE, to conduct comparisons with the rain gauge observations at multiple timescales (monthly, daily, and hourly). In addition, the two factors (elevation and precipitation intensity) relating to SPP performance are examined in this study. The main conclusions are as follows.

      SPPs can well capture the spatiotemporal characteristics of precipitation along the STR at multiple timescales, i.e., hourly, daily, monthly, and seasonal scales. Overall, SPPs can well reproduce the diurnal cycle of precipitation along the eastern STR, but the performance is not satisfactory in the western STR. Based on the evaluations, SPPs can be utilized to monitor the long-term (monthly and yearly scales) and short-term (daily and hourly scales) precipitation, as well as the extreme precipitation events in the meteorological and hydrological services for the STR construction and operation.

      The SPP with the best performance is different at various timescales, indicating that selecting SPP should refer to the SPP performance at particular timescales and in specific regions. Based on the evaluations of SPPs along the STR, we suggest that the GPM, TRMM, and GSMaP are suitable for the monthly timescale, and the GSMaP, GPM, and CMORPH are appropriate for the daily timescale. In addition, the FY2PRE, GPM, and GSMaP perform fairly well at hourly timescale. If there is only one choice for multiple timescales, we recommend the GPM, which has the best or second-best performance in most scenarios. Moreover, the record length and the spatiotemporal resolution of GPM are adequate for most meteorological and hydrological service needs.

      The SPP performance is strongly affected by the elevation and the precipitation intensity. In general, SPPs perform the worst at higher terrain and in the case of stronger precipitation.

      As the successor of TRMM, GPM has been remarkably improved in quality compared with TRMM and CMORPH. The performance of GSMaP is comparable to that of GPM. However, GSMaP performs substantially worse than GPM in characterizing light rain, whether for the daily timescale or the hourly timescale. The poor performance of PERSIANN compared with FY2PRE that also uses the infrared cloud image for retrieval indicates that the precipitation estimation from satellite observations through statistical methods (e.g., artificial neural networks) needs to be closely combined with the in situ ground-based observations, so as to obtain better results.

      The performance of various SPPs at multiple timescales is evaluated in the present study, despite within the specific region along the STR. It has the potential to provide some valuable feedback to algorithm developers, particularly for regions with complex terrain. It would be a useful step to further investigate the capability of SPPs in detecting and characterizing individual daily extreme precipitation events and hourly extreme precipitation events, and to examine the utilization of the SPPs as the calibration to improve the numerical weather prediction models along the STR and other complex mountainous regions.

Reference (65)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return