Advanced Search
Article Contents

Water Vapor Retrievals from Near-infrared Channels of the Advanced Medium Resolution Spectral Imager Instrument onboard the Fengyun-3D Satellite

Fund Project:

This research was funded by the National Key R&D Program of China (Grant Nos. 2018YFB0504900, 2018YFB0504901 and 2018YFB0504802) and the National Natural Science Foundation of China (Grant Nos. 41871249 and 41675036). The authors thank SuomiNet for providing the GPS data, as well as the FengYun Satellite Remote Sensing Data Service Network for providing the FY-3D/MERSI-II data


doi: 10.1007/s00376-020-0174-8

  • Water vapor plays a key role in weather, climate and environmental research on local and global scales. Knowledge about atmospheric water vapor and its spatiotemporal variability is essential for climate and weather research. Because of the advantage of a unique temporal and spatial resolution, satellite observations provide global or regional water vapor distributions. The advanced Medium Resolution Spectral Imager (MERSI) instrument—that is, MERSI-II—onboard the Fengyun-3D (FY-3D) meteorological satellite, has been one of the major satellite sensors routinely providing precipitable water vapor (PWV) products to the community using near-infrared (NIR) measurements since June 2018. In this paper, the major updates related to the production of the NIR PWV products of MERSI-II are discussed for the first time. In addition, the water vapor retrieval algorithm based on the MERSI-II NIR channels is introduced and derivations are made over clear land areas, clouds, and sun-glint areas over the ocean. Finally, the status and samples of the MERSI-II PWV products are presented. The accuracy of MERSI-II PWV products is validated using ground-based GPS measurements. The results show that the accuracies of the water vapor products based on the updated MERSI-II instrument are significantly improved compared with those of MERSI, because MERSI-II provides a better channel setting and new calibration method. The root-mean-square error and relative bias of MERSI-II PWV products are typically 1.8–5.5 mm and −3.0% to −14.3%, respectively, and thus comparable with those of other global remote sensing products of the same type.
    摘要: 大气水汽在区域和全球范围的天气、气候和环境研究中起着关键作用。了解大气水汽含量及其时空变化对于气候和天气研究至关重要。卫星遥感因其具有独特的时空分辨率的优势,可提供全球或区域性的水汽分布。我国风云3D(FY-3D)气象卫星上搭载的新一代的中分辨率光谱成像仪(即MERSI-II),与NASA的MODIS属于同类型传感器,具备基于近红外通道观测反演大气可降水产品(PWV)的能力。自2018年6月以来,MERSI-II可向用户业务提供全球范围的日、旬、月的大气可降水产品。本文首先重点讨论了MERSI-II大气可降水产品反演算法中涉及的主要改进和更新。继之,介绍了MERSI-II大气可降水产品的现状及产品示例。最后,基于地面的全球定位系统GPS的水汽观测数据,对MERSI-II PWV产品的精度进行了比对验证。结果表明,MERSI-II PWV产品的均方根误差和相对偏差通常分别为1.8~5.5 mm和−3.0%~−14.3%,与国际上同类型遥感仪器的水汽产品精度相当。与第一代MERSI相比,其水汽产品的精度有了显著提高主要与MERSI-II优化了通道设置,采用了新的辐射定标方法等因素有关。
  • 加载中
  • Figure 1.  Positions and widths of the five NIR channels used in MERSI (thick blue horizontal bars), MERSI-II (thick magenta horizontal bars), and MODIS (thick grey horizontal bars), and the two-way atmospheric water vapor transmittance spectra simulated for the midlatitude summer (MS) and midlatitude winter (MW) models in MODTRAN4.3 with a solar zenith angle of 45° and a nadir-looking geometry.

    Figure 2.  Examples of the transmittances simulated for the selected water vapor remote sensing channels in the 800–1100 nm range as a function of the total water vapor content in the sun–surface–sensor path: (a) FY3A/MERSI; (b) FY3D/MERSI-II.

    Figure 3.  RB in the calibration of the pre-launch calibration coefficients for the five bands used for the PWV retrieval. The red and blue bars denote the results from the WD-IVC method and the green bar denotes the results from the field VC campaign over Dunhuang.

    Figure 4.  Long-term calibration accuracy of the five channels used for the PWV retrieval between January 2018 and December 2019. The wavelength for Bands 15–19 are 865 nm, 905 nm, 936 nm, 940 nm and 1030 nm, respectively. (a) Window channels of bands 15 and 19. (b) Absorption channels of bands 16–18.

    Figure 5.  Normalized weighting factors as a function of the total PWV for the three absorption channels of MERSI-II.

    Figure 6.  Scatterplots of the reference water vapor values and those retrieved when the water absorption channel has a positive calibration bias of 5%, 10% and 15%: (a) 905 nm; (b) 936 nm; and (c) 940 nm. The gray dashed line has a 1:1 ratio and the blue line represents the PWV values for the water absorption channel with a positive calibration bias of 5% with respect to the reference PWV. The green and red lines represent the cases with a calibration bias of 10% and 15%, respectively.

    Figure 7.  Scatterplots between the water vapor values measured with GPS and those retrieved from MERSI-II NIR channels for the period 24–29 May 2018 globally: (a–d) pre-launch calibration coefficients used to calibrate MERSI-II L1B data; (e–h) on-orbit operational calibration coefficients used to calibrate MERSI-II Level-1B data. The gray dashed line has a 1:1 ratio.

    Figure 8.  (a) Example of a MERSI-II Level-2 PWV image over the Arabian Peninsula obtained at 1000 UTC 13 September 2019 and (b) the corresponding true color image composite processed from the MERSI-II Level-1B radiance of channels centered at 650 nm (red), 550 nm (green) and 470 nm (blue). There are no PWV retrievals over the water bodies, except for the sun-glint region of the ocean.

    Figure 10.  Monthly averaged global Level-3 water vapor images derived from FY3D MERSI-II NIR channel observations in (a) July 2019, (b) October 2019, and (c) December 2019.

    Figure 9.  Daily global Level-3 water vapor images derived from FY3D MERSI-II NIR channels on 2 July 2019.

    Figure 11.  (a) Scatterplot of the water vapor values measured with the GPS and MERSI-II NIR PWV values measured on clear days from January to December 2019 over North America. (b) Histogram of the difference between the GPS-measured and MERSI-II-retrieved PWVs using different channels.

    Figure 12.  (a) Monthly mean RB and (b) RMSE of the MERSI-II NIR PWV products compared with the GPS measurements from January to December 2019 over North America.

    Table 1.  Positions and widths of five NIR channels used for the water vapor monitoring of MERSI, MERSI-II and MODIS.

    MODISMERSIMERSI-II
    Position (nm)Width (nm)Position (nm)Width (nm)Position (nm)Width (nm)
    Window channel858358652086520
    124020103020103020
    Absorption channel905309052090520
    936109402093620
    940509802094050
    DownLoad: CSV

    Table 2.  List of pre-launch and on-orbit operational calibration coefficients.

    Band No.Center wavelength (nm)Pre-launchOn-orbit
    offsetslopeoffsetslope
    15*865−1.2990.00863−1.2990.00863
    16905−4.7060.02717−4.56050.0261
    17936−6.5070.02745−5.66940.0242
    18*940−6.7840.0286−6.7840.0286
    191030−4.0520.0295−6.04290.0335
    Note: *The on-orbit operational calibration coefficients for channels 15 and 18 are from the pre-launch test. The on-orbit operational calibration coefficients for channels 16, 17 and 19 are derived based on the WD-IVC method.
    DownLoad: CSV

    Table 3.  Selection of the transmittance–water vapor lookup table according to the solar zenith angle (SZA) and surface temperature (T).

    ConditionAtmospheric model usedConditionAtmospheric model used
    SZA ≤ 20Tropical60 < SZA ≤ 70Subarctic summer
    20 < SZA ≤ 45Midlatitude SummerSZA > 70, T < T0Subarctic winter
    45 < SZA ≤ 60Midlatitude WinterSZA > 70, TT0U.S. standard
    DownLoad: CSV

    Table 4.  RB of the calibration and water vapor transmittance for the pre-launch and on-orbit operational calibration coefficients.

    Band No.Center wavelength (nm)RB of the calibration (%)RB of the transmittance (%)
    Two-channel ratioThree-channel ratio
    15*8650.00
    169056.586.589.15
    1793611.4911.4916.35
    18*9400.000.007.49
    191030−9.68
    Note: *For channels 15 and 18, the pre-launch test results were used as the on-orbit operational calibration coefficients. Hence, there is no difference in the calibration results between the pre-launch and on-orbit operational calibration coefficients. For the two-channel ratio technique, the window channel with the shorter wavelength (865 nm) was adopted to calculate the PWV transmittance.
    DownLoad: CSV
  • Alshawaf, F., T. Fuhrmann, A. Knöpfler, X. Luo, M. Mayer, S. Hinz, and B. Heck, 2015: Accurate estimation of atmospheric water vapor using GNSS observations and surface meteorological data. IEEE Trans. Geosci. Remote Sens., 53, 3764−3771, https://doi.org/10.1109/TGRS.2014.2382713.
    Aumann, H. H., and Coauthors, 2003: AIRS/AMSU/HSB on the aqua mission: design, science objectives, data products, and processing systems. IEEE Trans. Geosci. Remote Sens., 41, 253−264, https://doi.org/10.1109/TGRS.2002.808356.
    Berk, A., L. S. Bernstein, and D. C. Robertson, 1987: MODTRAN: A moderate resolution model for LOWTRAN. Technical Report, AD-A-185384/5/XAB.
    Bevis, M., S. Businger, T. A. Herring, C. Rocken, R. A. Anthes, and R. H. Ware, 1992: GPS meteorology: Remote sensing of atmospheric water vapor using the global positioning system. J. Geophys. Res., 97, 15 787−15 801, https://doi.org/10.1029/92JD01517.
    Chen, L., X. Q. Hu, N. Q. Xu, and P. Zhang, 2013: The application of deep convective clouds in the calibration and response monitoring of the reflective solar bands of FY-3A/MERSI (Medium Resolution Spectral Imager). Remote Sensing, 5(12), 6958−6975, https://doi.org/10.3390/rs5126958.
    Chen, L., P. Zhang, J. Y. Lv, N. Xu, and X. Q. Hu, 2017: Radiometric calibration evaluation for RSBs of Suomi-NPP/VIIRS and Aqua/MODIS based on the 2015 Dunhuang Chinese Radiometric Calibration Site in situ measurements. Int. J. Remote Sens., 38(20), 5640−5656, https://doi.org/10.1080/01431161.2017.1343514.
    Czajkowski, K. P., S. N. Goward, D. Shirey, and A. Walz, 2002: Thermal remote sensing of near-surface water vapor. Remote Sens. Environ., 79, 253−265, https://doi.org/10.1016/S0034-4257(01)00277-2.
    Dessler, A. E., Z. Zhang, and P. Yang, 2008: Water-vapor climate feedback inferred from climate fluctuations, 2003-2008. Geophys. Res. Lett., 35, L20704, https://doi.org/10.1029/2008GL035333.
    Diedrich, H., R. Preusker, R. Lindstrot, and J. Fischer, 2015: Retrieval of daytime total columnar water vapour from MODIS measurements over land surfaces. Atmospheric Measurement Techniques, 8, 823−836, https://doi.org/10.5194/amt-8-823-2015.
    Elgered, G., H. P. Plag, H. van der Marel, S. Barlag, and J. Nash, 2005: Ground-based GPS for operational numerical weather prediction and climate applications. Final Report, COST Action 716 EUR 21639 edition, COST Office, Luxembourg. [Available online athttps://core.ac.uk/download/pdf/70606169.pdf]
    Fix, A., G. Poberaj, C. Kiemle, H. Flentje, R. Busen, M. Fiebig, and G. Ehret, 2002: MIPAS Validation with the DLR Falcon. Proc. Envisat Validation Workshop, H. Lacoste, Ed., ESA SP-531, Frascati, Italy, 721pp.
    Gao, B. C., and Y. J. Kaufman, 2003: Water vapor retrievals using Moderate Resolution Imaging Spectroradiometer (MODIS) near-infrared channels. J. Geophys. Res., 108, 4389, https://doi.org/10.1029/2002JD003023.
    Gong, S. Q., D. F. Hagan, J. Lu, and G. J. Wang, 2018: Validation on MERSI/FY-3A precipitable water vapor product. Advances in Space Research, 61, 413−425, https://doi.org/10.1016/j.asr.2017.10.005.
    Held, I. M., and B. J. Soden, 2000: Water vapor feedback and global warming. Annual Review of Energy and the Environment, 25, 441−475, https://doi.org/10.1146/annurev.energy.25.1.441.
    Hu, X. Q., Y. F. Huang, Q. F. Lu, and J. Zheng, 2011: Retrieving precipitable water vapor based on the near-infrared data of FY-3A satellite. Journal of Applied Meteorological Science, 22, 46−56, https://doi.org/10.3969/j.issn.1001-7313.2011.01.005. (in Chinese with English abstract
    Kaufman, Y. J., and B. C. Gao, 1992: Remote sensing of water vapor in the near IR from EOS/MODIS. IEEE Trans. Geosci. Remote Sens., 30, 871−884, https://doi.org/10.1109/36.175321.
    Li, Y., and Coauthors, 2009: Post launch site calibration of visible and near-infrared channels of FY-3A visible and infrared radiometers. Optics and Precision Engineering, 17, 2966−2973, https://doi.org/10.3321/j.issn:1004-924X.2009.12.014.
    Li, Z. H., J. P. Muller, and P. Cross, 2003: Comparison of precipitable water vapor derived from radiosonde, GPS, and moderate-resolution imaging spectroradiometer measurements. J. Geophys. Res. Atmos., 108, 4651, https://doi.org/10.1029/2003JD003372.
    Li, Z. L., B. H. Tang, H. Wu, H. Z. Ren, G. J. Yan, Z. M. Wan, I. F. Trigo, and J. A. Sobrino, 2013: Satellite-derived land surface temperature: current status and perspectives. Remote Sens. Environ., 131, 14−37, https://doi.org/10.1016/j.rse.2012.12.008.
    Liu, Z. Z., M. S. Wong, J. Nichol, and P. W. Chan, 2013: A multi-sensor study of water vapour from radiosonde, MODIS and AERONET: A case study of Hong Kong. International Journal of Climatology, 33, 109−120, https://doi.org/10.1002/joc.3412.
    Liu, Z. Z., B. Y. Chen, S. T. Chan, Y. C. Cao, Y. Gao, K. F. Zhang, and J. Nichol, 2015a: Analysis and modelling of water vapour and temperature changes in Hong Kong using a 40-year radiosonde record: 1973−2012. International Journal of Climatology, 35, 462−474, https://doi.org/10.1002/joc.4001.
    Liu, H. L., S. H. Tang, S. L. Zhang, and J. Y. Hu, 2015b: Evaluation of MODIS water vapour products over China using radiosonde data. Int. J. Remote Sens., 36, 680−690, https://doi.org/10.1080/01431161.2014.999884.
    Liu, H. L., S. H. Tang, J. Y. Hu, S. L. Zhang, and X. B. Deng, 2017: An improved physical split-window algorithm for precipitable water vapor retrieval exploiting the water vapor channel observations. Remote Sens. Environ., 194, 366−378, https://doi.org/10.1016/j.rse.2017.03.031.
    Liu, J. M., H. Liang, Z. Sun, and X. Zhou, 2006: Validation of the moderate-resolution imaging Spectroradiometer precipitable water vapor product using measurements from GPS on the Tibetan Plateau. J. Geophys. Res., 111(D14), D14103, https://doi.org/10.1029/2005JD007028.
    Mockler, S. B., 1995: Water vapor in the climate system. American Geophysical Union, Washington, DC. [Available online at https://www.eso.org/gen-fac/pubs/astclim/espas/pwv/mockler.html]
    Park, C. G., J. H. Baek, and J. H. Cho, 2010: Analysis on characteristics of radiosonde bias using GPS precipitable water vapor. Journal of Astronomy and Space Sciences, 27(3), 213−220, https://doi.org/10.5140/jass.2010.27.3.213.
    Prasad, A. K., and R. P. Singh, 2009: Validation of MODIS Terra, AIRS, NCEP/DOE AMIP-II Reanalysis-2, and AERONET Sun photometer derived integrated precipitable water vapor using ground-based GPS receivers over India. J. Geophys. Res., 114, D05107, https://doi.org/10.1029/2008JD011230.
    Qin, Z., A. Karnieli, and P. Berliner, 2001: A mono-window algorithm for retrieving land surface temperature from Landsat TM data and its application to the Israel-Egypt border region. Int. J. Remote Sens., 22, 3719−3746, https://doi.org/10.1080/01431160010006971.
    Sun, L., and Coauthors, 2012a: On-orbit response variation analysis of FY-3 MERSI reflective solar bands based on Dunhuang site calibration. Spectroscopy and Spectral Analysis, 32, 1869−1877, https://doi.org/10.3964/j.issn.1000-0593(2012)07-1869-09.
    Sun, L., X. Q. Hu, M. H. Guo, and N. Xu, 2012b: Multisite calibration tracking for FY-3A MERSI solar bands. IEEE Trans. Geosci. Remote Sens., 50(12), 4929−4942, https://doi.org/10.1109/TGRS.2012.2215613.
    Tregoning, P., R. Boers, D. O’Brien, and M. Hendy, 1998: Accuracy of absolute precipitable water vapor estimates from GPS observations. J. Geophys. Res., 103, 28 701−28 710, https://doi.org/10.1029/98JD02516.
    Vermote, E. F., N. Z. El Saleous, and C. O. Justice, 2002: Atmospheric correction of MODIS data in the visible to middle infrared: First results. Remote Sens. Environ., 83, 97−111, https://doi.org/10.1016/S0034-4257(02)00089-5.
    Vermote, E. F., D. Tanre, J. L. Deuze, M. Herman, and J. J. Morcette, 1997: Second Simulation of the Satellite Signal in the Solar Spectrum, 6S: An overview. IEEE Trans. Geosci. Remote Sens., 35(3), 675−686, https://doi.org/10.1109/36.581987.
    Wang, H., X. Liu, K. Chance, G. G. Abad, and C. C. Miller, 2014: Water vapor retrieval from OMI visible spectra. Atmospheric Measurement Techniques, 7(6), 1901−1913, https://doi.org/10.5194/amt-7-1901-2014.
    Wang, L., X. Q. Hu, and L. Chen, 2015: FY-3C/MERSI caliration for solar band using multi-reflectance stable targets. Optics and Precision Engineering, 23(7), 1911−1920, https://doi.org/10.3788/OPE.20152307.1911. (in Chinese with English abstract
    Wang, L., X. Q. Hu, and L. Chen, 2017: Wide dynamic nonlinear radiometric calibration of optical satellite sensors using multiple stable earth targets. Journal of Remote Sensing, 21(6), 892−906, https://doi.org/10.11834/jrs.20176351. (in Chinese with English abstract
    Ware, R. H., and Coauthors, 2000: SuomiNet: A real-time national GPS network for atmospheric research and education. Bull. Amer. Meteor. Soc., 81, 677−694, https://doi.org/10.1175/1520-0477(2000)081<0677:SARNGN>2.3.CO;2.
    Wong, M. S., X. M. Jin, Z. Z. Liu, J. Nichol, and P. W. Chan, 2015: Multi-sensors study of precipitable water vapour over mainland China. International Journal of Climatology, 35, 3146−3159, https://doi.org/10.1002/joc.4199.
    Xu, N., R. H. Wu, X. Q. Hu, L. Chen, L. Wang, and L. Sun, 2015: Integrated method for on-obit wide dynamic vicarious calibration of FY-3C MERSI reflective solar bands. Acta Optica Sinica, 35(12), 1228001, https://doi.org/10.3788/AOS201535.1228001. (in Chinese with English abstract
    Xu, N., and Coauthors, 2018: Prelaunch calibration and radiometric performance of the advanced MERSI II on FengYun-3D. IEEE Trans. Geosci. Remote Sens., 56, 4866−4875, https://doi.org/10.1109/TGRS.2018.2841827.
    Yang, Z. D., and Coauthors, 2019: Capability of Fengyun-3D satellite in earth system observation. J. Meteor. Res., 33, 1113−1130, https://doi.org/10.1007/s13351-019-9063-4.
    Zhang, P., and Coauthors, 2019: Latest progress of the Chinese meteorological satellite program and core data processing technologies. Adv. Atmos. Sci., 36, 1027−1045, https://doi.org/10.1007/s00376-019-8215-x.
    Zhao, T. B., A. G. Dai, and J. H. Wang, 2012: Trends in tropospheric humidity from 1970 to 2008 over China from a homogenized radiosonde dataset. J. Climate, 25, 4549−4567, https://doi.org/10.1175/jcli-d-11-00557.1.
  • [1] ZHANG Meng, NI Yunqi, ZHANG Fuqing, 2007: Variational Assimilation of GPS Precipitable Water Vapor and Hourly Rainfall Observations for a Meso- Scale Heavy Precipitation Event During the 2002 Mei-Yu Season, ADVANCES IN ATMOSPHERIC SCIENCES, 24, 509-526.  doi: 10.1007/s00376-007-0509-8
    [2] BI Yanmeng, MAO Jietai, LI Chengcai, 2006: Preliminary Results of 4-D Water Vapor Tomography in the Troposphere Using GPS, ADVANCES IN ATMOSPHERIC SCIENCES, 23, 551-560.  doi: 10.1007/s00376-006-0551-y
    [3] Houaria NAMAOUI, Salem KAHLOUCHE, Ahmed Hafid BELBACHIR, Roeland Van MALDEREN, Hugues BRENOT, Eric POTTIAUX, 2017: GPS Water Vapor and Its Comparison with Radiosonde and ERA-Interim Data in Algeria, ADVANCES IN ATMOSPHERIC SCIENCES, 34, 623-634.  doi: 10.1007/s00376-016-6111-1
    [4] WANG Xiaoying, WANG Xianliang, DAI Ziqiang, KE Fuyang, CAO Yunchang, WANG Feifei, SONG Lianchun, 2014: Tropospheric Wet Refractivity Tomography Based on the BeiDou Satellite System, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 355-362.  doi: 10.1007/s00376-013-2311-0
    [5] Ha-Taek KWON, Eui-Hyun JUNG, Gyu-Ho LIM, 2010: A Comparison of GPS- and NWP-derived PW Data over the Korean Peninsula, ADVANCES IN ATMOSPHERIC SCIENCES, 27, 871-882.  doi: 10.1007/s00376-009-9069-4
    [6] XU Xiaohua, LUO Jia, SHI Chuang, 2009: Comparison of COSMIC Radio Occultation Refractivity Profiles with Radiosonde Measurements, ADVANCES IN ATMOSPHERIC SCIENCES, 26, 1137-1145.  doi: 10.1007/s00376-009-8066-y
    [7] CAO Yunchang, CHEN Yongqi, LI Pingwha, 2006: Wet Refractivity Tomography with an Improved Kalman-Filter Method, ADVANCES IN ATMOSPHERIC SCIENCES, 23, 693-699.  doi: 10.1007/s00376-006-0693-y
    [8] DING Jincai, YANG Yinming, YE Qixin, HUANG Yan, MA Xiaoxing, MA Leiming, Y. R. GUO, 2007: Moisture Analysis of a Squall Line Case Based on Precipitable Water Vapor Data from a Ground-Based GPS Network in the Yangtze River Delta, ADVANCES IN ATMOSPHERIC SCIENCES, 24, 409-420.  doi: 10.1007/s00376-007-0409-y
    [9] Minqiang ZHOU, Zhili DENG, Charles ROBERT, Xingying ZHANG, Lu ZHANG, Yapeng WANG, Chengli QI, Pucai WANG, Martine De MAZIÈRE, 2024: The First Global Map of Atmospheric Ammonia (NH3) as Observed by the HIRAS/FY-3D Satellite, ADVANCES IN ATMOSPHERIC SCIENCES, 41, 379-390.  doi: 10.1007/s00376-023-3059-9
    [10] Luyao QIN, Yaodeng CHEN, Gang MA, Fuzhong WENG, Deming MENG, Peng ZHANG, 2023: Assimilation of FY-3D MWTS-II Radiance with 3D Precipitation Detection and the Impacts on Typhoon Forecasts, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 900-919.  doi: 10.1007/s00376-022-1252-x
    [11] Eun-Han KWON, Jinlong LI, B. J. SOHN, Elisabeth WEISZ, 2012: Use of Total Precipitable Water Classification of A Priori Error and Quality Control in Atmospheric Temperature and Water Vapor Sounding Retrieval, ADVANCES IN ATMOSPHERIC SCIENCES, 29, 263-273.  doi: 10.1007/s00376-011-1119-z
    [12] Li Shuyong, Wang Bin, Zhang Xin, 2001: The Parallel Computing of GPS Ray-shooting Model, ADVANCES IN ATMOSPHERIC SCIENCES, 18, 1185-1191.  doi: 10.1007/s00376-001-0032-2
    [13] KUANG Zheng, WANG Bin, YANG Hualin, 2003: A Rapid Optimization Algorithm for GPS Data Assimilation, ADVANCES IN ATMOSPHERIC SCIENCES, 20, 437-441.  doi: 10.1007/BF02690801
    [14] WANG Yunfeng, WANG Bin, 2003: The Variational Assimilation Experiment of GPS Bending Angle, ADVANCES IN ATMOSPHERIC SCIENCES, 20, 479-486.  doi: 10.1007/BF02690806
    [15] YANG Lu, WANG Zhenhui, CHU Yanli, ZHAO Hang, TANG Min, 2014: Water Vapor Motion Signal Extraction from FY-2E Longwave Infrared Window Images for Cloud-free Regions: The Temporal Difference Technique, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 1386-1394.  doi: 10.1007/s00376-014-3165-9
    [16] YAO Zhanyu, XU Chenhai, YUAN Jian, LI Wanbiao, ZHU Yuanjing, ZHAO Bolin, 2003: Estimations of Precipitable Water and Its Characteristics during the HUBEX/IOP 1998, ADVANCES IN ATMOSPHERIC SCIENCES, 20, 219-226.  doi: 10.1007/s00376-003-0007-6
    [17] B. J. SOHN, 2024: Water Vapor Information from Satellite and Its Applications, ADVANCES IN ATMOSPHERIC SCIENCES, 41, 188-189.  doi: 10.1007/s00376-023-3008-7
    [18] Wei Chong, Xue Yongkang, Zhu Xiaoming, Zou Shouxiang, 1984: DETERMINATION OF ATMOSPHERIC PRECIPITABLE WATER AND HUMIDITY PROFILES BY A GROUND-BASED 1.35 cm RADIOMETER, ADVANCES IN ATMOSPHERIC SCIENCES, 1, 119-139.  doi: 10.1007/BF03187623
    [19] ZHANG Xin, LIU Yuewei, WANG Bin, JI Zhongzhen, 2004: Parallel Computing of a Variational Data Assimilation Model for GPS/MET Observation Using the Ray-Tracing Method, ADVANCES IN ATMOSPHERIC SCIENCES, 21, 220-226.  doi: 10.1007/BF02915708
    [20] WU Xue*, WANG Xin, and LÜ Daren, 2014: Retrieval of Vertical Distribution of Tropospheric Refractivity through Ground-Based GPS Observation, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 37-47.  doi: 10.1007/s00376-013-2215-z

Get Citation+

Export:  

Share Article

Manuscript History

Manuscript received: 09 June 2020
Manuscript revised: 02 September 2020
Manuscript accepted: 07 September 2020
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

Water Vapor Retrievals from Near-infrared Channels of the Advanced Medium Resolution Spectral Imager Instrument onboard the Fengyun-3D Satellite

    Corresponding author: Ling WANG, lingw@cma.cn
  • 1. Key Laboratory of Radiometric Calibration and Validation for Environmental Satellites, China Meteorological Administration, Beijing 100081, China
  • 2. National Satellite Meteorological Center, China Meteorological Administration, Beijing 100081, China

Abstract: Water vapor plays a key role in weather, climate and environmental research on local and global scales. Knowledge about atmospheric water vapor and its spatiotemporal variability is essential for climate and weather research. Because of the advantage of a unique temporal and spatial resolution, satellite observations provide global or regional water vapor distributions. The advanced Medium Resolution Spectral Imager (MERSI) instrument—that is, MERSI-II—onboard the Fengyun-3D (FY-3D) meteorological satellite, has been one of the major satellite sensors routinely providing precipitable water vapor (PWV) products to the community using near-infrared (NIR) measurements since June 2018. In this paper, the major updates related to the production of the NIR PWV products of MERSI-II are discussed for the first time. In addition, the water vapor retrieval algorithm based on the MERSI-II NIR channels is introduced and derivations are made over clear land areas, clouds, and sun-glint areas over the ocean. Finally, the status and samples of the MERSI-II PWV products are presented. The accuracy of MERSI-II PWV products is validated using ground-based GPS measurements. The results show that the accuracies of the water vapor products based on the updated MERSI-II instrument are significantly improved compared with those of MERSI, because MERSI-II provides a better channel setting and new calibration method. The root-mean-square error and relative bias of MERSI-II PWV products are typically 1.8–5.5 mm and −3.0% to −14.3%, respectively, and thus comparable with those of other global remote sensing products of the same type.

摘要: 大气水汽在区域和全球范围的天气、气候和环境研究中起着关键作用。了解大气水汽含量及其时空变化对于气候和天气研究至关重要。卫星遥感因其具有独特的时空分辨率的优势,可提供全球或区域性的水汽分布。我国风云3D(FY-3D)气象卫星上搭载的新一代的中分辨率光谱成像仪(即MERSI-II),与NASA的MODIS属于同类型传感器,具备基于近红外通道观测反演大气可降水产品(PWV)的能力。自2018年6月以来,MERSI-II可向用户业务提供全球范围的日、旬、月的大气可降水产品。本文首先重点讨论了MERSI-II大气可降水产品反演算法中涉及的主要改进和更新。继之,介绍了MERSI-II大气可降水产品的现状及产品示例。最后,基于地面的全球定位系统GPS的水汽观测数据,对MERSI-II PWV产品的精度进行了比对验证。结果表明,MERSI-II PWV产品的均方根误差和相对偏差通常分别为1.8~5.5 mm和−3.0%~−14.3%,与国际上同类型遥感仪器的水汽产品精度相当。与第一代MERSI相比,其水汽产品的精度有了显著提高主要与MERSI-II优化了通道设置,采用了新的辐射定标方法等因素有关。

1.   Introduction
  • Water vapor is the most abundant greenhouse gas in the Earth’s atmosphere and its movement accounts for most of the atmospheric energy transport, affecting weather systems and the climate in the short and long term, respectively (Held and Soden, 2000; Dessler et al., 2008; Liu et al., 2015a, 2017). Furthermore, water vapor plays an important role in cloud formation, hydrological and energy cycles, and atmospheric chemistry and dynamics (Fix et al., 2002; Park et al., 2010; Zhao et al., 2012). Moreover, water vapor affects ground- or space-based remote sensing measurements. For example, the signals of the Global Navigation System/Global Positioning System (GNSS/GPS) are greatly affected by water vapor in the atmosphere (Bevis et al., 1992). It is among the important geophysical parameters that affects surface remote sensing applications, such as land surface temperature retrieval and atmospheric correction of satellite data (Qin et al., 2001; Vermote et al., 2002; Li et al., 2013; Liu et al., 2017). However, water vapor fields vary significantly across the globe, with values ranging from ~5 cm near the equator to less than one tenth at the poles (Mockler, 1995). The quantification of water vapor at high spatiotemporal resolution is important in this scientific field. The precipitable water vapor (PWV) is an important parameter to account for the total atmospheric water vapor in a vertical column of a cross-sectional area extending from the Earth’s surface to the top of the atmosphere (TOA).

    Numerous techniques have been used to obtain the PWV, such as radiosonde, GPS, ground-based sun photometers, ground-based microwave radiometers, and satellite observations (Bevis et al., 1992; Czajkowski et al., 2002; Li et al., 2003; Gao and Kaufman, 2003; Alshawaf et al., 2015). Because of their unique temporal and spatial resolution, satellite observations can be used to determine global or regional PWV distributions (Kaufman and Gao, 1992; Gao and Kaufman, 2003; Liu et al., 2017). They have become the most widely used technique in water vapor monitoring (Aumann et al., 2003). Sensors can be divided into four categories based on the propagated signals: visible (VIS), near-infrared (NIR), infrared (IR), and microwave measurements (Wang et al., 2014). The Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra and Aqua, the Medium Resolution Imaging Spectrometer (MERIS) onboard Envisat, and MERSI/MERSI-II onboard the Fengyun-3 (FY-3) meteorological satellite series, are the three major sensors providing NIR-based PWV products to the community. Many studies have been carried out to validate and improve the accuracy of the PWV products of MODIS and MERSI (Prasad and Singh, 2009; Liu et al., 2013; Diedrich et al., 2015; Wong et al., 2015; Liu et al., 2015b). However, few researchers focused on the efficiency of MERSI’s PWV products. In this study, we therefore introduce the current status of FY-3D/MERSI-II’s PWV products in detail, including updates of the MERSI-II instrument regarding the NIR PWV retrieval, water vapor retrieval algorithm using MERSI-II NIR channels, and product samples as well as product accuracy.

    An overview of the FY-3D/MERSI-II instrument is provided in section 2, and the changes in the NIR channel settings as well as calibration methods used for the PWV retrieval of MERSI and MERSI-II are discussed. The PWV retrieval technique and validation methods based on ground-based GPS measurements are introduced in section 3. The status of the PWV products of MERSI-II as well as the validated results are illustrated in section 4. Moreover, several issues related to the accuracy of MERSI-II’s PWV products are discussed in section 4. Finally, conclusions are drawn in section 5.

2.   Instrument overview and improvements
  • The FY-3 series of satellites are the second-generation of Chinese polar-orbiting, sun-synchronous meteorological satellites. FY-3D is the fourth and latest satellite, which was successfully launched on 15 November 2017. The equator crossing time is 13:40 local time in the rising intersection node and the average altitude of the orbit is 830.73 km (Yang et al., 2019; Zhang et al., 2019). In contrast to the previous FY-3A/B/C satellites, the FY-3D satellite was designed to be equipped with an advanced version of MERSI (MERSI-II). The previous version of MERSI onboard the FY-3A/B/C satellites included 20 bands: 19 reflective solar bands (RSBs) over the spectral range of 0.4–2.1 μm, and one thermal emissive band (TEB) over the range of 10–12.5 μm (Sun et al., 2012a). The updated MERSI-II sensor contains 25 bands—that is, 20 RSBs and 5 TEBs—covering the VIS (at 0.41 μm) to longwave IR (at 12.2 μm; Xu et al., 2018) spectrum. The first four RSBs (blue, green, red, NIR), and two TEB split windows, have a spatial resolution of 250 m at nadir, and the other bands have a spatial resolution of 1 km at nadir (Xu et al., 2018).

  • PWV is one of the major atmospheric products of MERSI-II. Two PWV products are provided by MERSI-II, which are derived from the NIR and IR channels, respectively. The NIR PWV products are derived via comparison between the reflected solar radiance in the absorption channels with those in nearby non-absorption channels in the range of 0.8–1.3 μm. The spectral positions and widths of these channels are listed in Table 1 and illustrated in Fig. 1. The previous MERSI and MODIS channel specifications of the NIR PWV products are also provided in Table 1 and Fig. 1.

    MODISMERSIMERSI-II
    Position (nm)Width (nm)Position (nm)Width (nm)Position (nm)Width (nm)
    Window channel858358652086520
    124020103020103020
    Absorption channel905309052090520
    936109402093620
    940509802094050

    Table 1.  Positions and widths of five NIR channels used for the water vapor monitoring of MERSI, MERSI-II and MODIS.

    Figure 1.  Positions and widths of the five NIR channels used in MERSI (thick blue horizontal bars), MERSI-II (thick magenta horizontal bars), and MODIS (thick grey horizontal bars), and the two-way atmospheric water vapor transmittance spectra simulated for the midlatitude summer (MS) and midlatitude winter (MW) models in MODTRAN4.3 with a solar zenith angle of 45° and a nadir-looking geometry.

    Three absorption channels and two non-absorption channels in the 0.8–1.3 μm range are included in the design of all three instruments to monitor atmospheric water vapor. The same non-absorption channels were selected for MERSI and MERSI-II, which are centered at 865 and 1030 nm and have an identical spectral width of 20 nm. However, different absorption channels were selected for the two MERSI instruments. Two strong water vapor absorption channels, both centered at 940 nm but with different widths (20 and 50 nm), were selected for MERSI-II, while only one strong water vapor absorption channel centered at 940 nm with a spectral width of 20 nm was used in MERSI. One weak absorption channel centered at 905 nm was used for MERSI-II, while two weak absorption channels at 905 and 980 nm were utilized in MERSI. The setup of the water vapor absorption channels of MERSI-II is similar to that of MODIS.

    The simulated transmittances of these selected channels are plotted in Fig. 2 for different instruments as a function of the total water vapor. A detailed simulation setup is presented in section 3. The transmittance of the absorption channels decreases with increasing atmospheric water vapor concentration. The transmittance of the strong absorption channel at 940 nm with a narrow spectral width (20 nm) decreases the fastest among the three absorption channels, when the water vapor content is small. This means that this channel is most sensitive to changes in the atmospheric water vapor concentration under dry conditions (PWV < 2 cm). By contrast, the transmittance of the weak absorption channel at 905 nm decreases the fastest among the three absorption channels, when the water vapor content is large. This means that this channel is most sensitive to changes in the atmospheric water vapor concentration under humid conditions (PWV > 4 cm). The same is true for the weak absorption channel at 980 nm, which is used in MERSI. To avoid redundant information, only one weak absorption channel centered at 905 nm was kept in MERSI-II and the 980 channel was replaced with a wide (spectral width of 50 nm) and strong absorption channel at 940 nm. The transmittance of this absorption channel varies between that of the weak absorption channel (905 nm) and that of the narrow strong absorption channel (940 nm, with 20 nm spectral width), which provides additional information for the water vapor retrieval under medium water vapor conditions. Hence, compared with the previous MERSI, MERSI-II provides richer information for the water vapor retrieval under different conditions (dry, intermediate, and humid).

    Figure 2.  Examples of the transmittances simulated for the selected water vapor remote sensing channels in the 800–1100 nm range as a function of the total water vapor content in the sun–surface–sensor path: (a) FY3A/MERSI; (b) FY3D/MERSI-II.

  • Because the water vapor concentration is derived from spectral radiance measurements of satellite sensors, the accuracy of the sensor calibration affects the accuracy of the retrieved water vapor content. A set of calibration coefficients for the RSBs was determined with a pre-launch calibration test under ambient laboratory conditions and in the field. However, the pre-launch calibration coefficients needed to be updated when the satellite was launched because the environment of the satellite instrument had changed. Hence, on-orbit calibration is performed using either onboard calibration systems or vicarious calibration (VC) during the orbital operation of the satellite instruments. Due to the lack of an effective onboard calibration, the operational calibration of the previous MERSI RSBs was mainly based on field VC campaigns conducted over a China Radiometric Calibration Site; namely, Dunhuang (Li et al., 2009). The reference radiance at the TOA used for this method was obtained from a radiative transfer model (RTM) and synchronous in-situ measurements. It has been proven that this method can yield an accuracy of 3%–5% for most of the RSBs, except for the water vapor absorption channels. The relatively large error of in-situ surface reflectance measurements in the water vapor absorption channels, uncertainty in the water vapor measurements, and relatively low accuracy of the RTM simulation in the water vapor absorption channels, are three major factors leading to the poor calibration accuracy of water vapor absorption channels. Furthermore, this method has disadvantages such as high cost, low calibration frequency, and limited radiance dynamic range based on the use of a single calibration site.

    The recently developed automatic calibration using multiple Global Pseudo-invariant Calibration Sites (PICSs), including desert, dry salt lake, and dark ocean sites, has gradually become another major operational method that can be used for FY satellite inflight calibration (Sun et al., 2012b; Wang et al., 2015, 2017). This method can achieve calibration at daily frequency. Furthermore, a broad radiance dynamic range is achieved by using multiple sites with different brightness values (Wang et al., 2017). Benefitting from the increase in the dynamic range of the calibration, the calibration uncertainty can be reduced (Wang et al., 2017). Similar to the VC in the field, the calibration reference of this method is obtained from an RTM such as 6SV (Vermote et al., 1997). Further, the calibration sites are located on the Earth’s surface and the accuracy of the estimation of atmospheric parameters (input parameters required for radiative calculation) also affects the accuracy of the results of the RTM. To improve the calibration accuracy of the water vapor absorption channels, a wide dynamic integrated VC (WD-IVC) method has been developed for MERSI instruments (after FY-3C; Xu et al., 2015). In addition to PICS, deep convective cloud (DCC) and lunar targets are included in this method. The spectral features of the DCCs are similar to those of the reference white board used for radiometric calibration. It provides a high and stable reflectance for calibration. The lunar irradiance reflected by the sun has a high photometric stability. Thus, the lunar target is always selected as both the radiometric standard for Earth-orbiting satellite-borne instrument calibration and Earth’s illumination source in the sky at night for ground-based and satellite remote sensing at night (Chen et al., 2013, 2017; Zhang et al., 2019). Another important feature is that DCCs are distributed above the troposphere and the moon has no atmosphere; thus, the effect of the atmosphere on the radiative calibration using these two targets is small.

    The WD-IVC and field VC campaign over Dunhuang during the on-orbit test (the first year after MERSI-II’s launch) showed that the calibration accuracy of the pre-launch calibration coefficients for the 905, 936, and 1030 nm bands was worse than 5%, failing to meet the calibration accuracy requirements (Fig. 3). For these three channels, new operational calibration coefficients were calculated using the WD-IVC method. Meanwhile, for the 865 and 940 nm channels, whose calibration accuracy meets the requirements of ±5%, the pre-launch calibration coefficients were used in the current operational system. As illustrated in Fig. 4, by using the current operational calibration coefficients obtained during the on-orbit test (last two columns in Table 2), the long-term calibration accuracy of these five channels for PWV retrieval has been maintained within ±5% since MERSI-II’s launch.

    Figure 3.  RB in the calibration of the pre-launch calibration coefficients for the five bands used for the PWV retrieval. The red and blue bars denote the results from the WD-IVC method and the green bar denotes the results from the field VC campaign over Dunhuang.

    Figure 4.  Long-term calibration accuracy of the five channels used for the PWV retrieval between January 2018 and December 2019. The wavelength for Bands 15–19 are 865 nm, 905 nm, 936 nm, 940 nm and 1030 nm, respectively. (a) Window channels of bands 15 and 19. (b) Absorption channels of bands 16–18.

    Band No.Center wavelength (nm)Pre-launchOn-orbit
    offsetslopeoffsetslope
    15*865−1.2990.00863−1.2990.00863
    16905−4.7060.02717−4.56050.0261
    17936−6.5070.02745−5.66940.0242
    18*940−6.7840.0286−6.7840.0286
    191030−4.0520.0295−6.04290.0335
    Note: *The on-orbit operational calibration coefficients for channels 15 and 18 are from the pre-launch test. The on-orbit operational calibration coefficients for channels 16, 17 and 19 are derived based on the WD-IVC method.

    Table 2.  List of pre-launch and on-orbit operational calibration coefficients.

3.   Water vapor retrieval algorithm and validation methods
  • The radiance at a downward-looking satellite sensor in NIR channels can be approximated as (Gao and Kaufman, 2003)

    where Lsensor(λ) is the radiance at the sensor channel with a wavelength of λ, Lpath(λ) is the path scattered radiance, μ0 is the cosine of the solar zenith angle, E0(λ) is the solar flux at the TOA, T(λ) is the total atmospheric transmittance, and ρ(λ) is the surface bidirectional reflectance. Because the aerosol optical thickness is small in the NIR region, Lpath(λ) can be ignored. Hence, Eq. (1) can be written as

    where ${\rho }^{\mathrm{*}}\left(\lambda \right) $ is πLsensor(λ)/μ0E0(λ), which is defined as apparent reflectance (Gao and Kaufman, 2003). The term T(λ) contains information about the total amount of water vapor in the sun–surface–sensor path. However, the surface reflectance term of (λ) in Eq. (2) is unknown. The reflectance values at a given wavelength differ for different types of surfaces. Therefore, it is impossible to directly obtain the water vapor transmittances from radiances of individual absorption channels. A channel ratio technique has been proposed to solve this problem. Specifically, if the surface reflectance remains constant with changing wavelength, a two-channel ratio of an absorption channel along with a window channel can be used to estimate the water vapor content according to the transmittance of the absorption channel. The transmittance at the water vapor absorption channel can be written as (Gao and Kaufman, 2003; Hu et al., 2011)

    If the surface reflectance linearly correlates with the wavelength, a three-channel ratio of the absorption channel with a combination of two window channels will be used to obtain the transmittance of the absorption channel. The transmittance at the water vapor absorption channel can be written as

    where the subscript “wv” denotes the water vapor channel; the subscripts “01” and “02” denote the window channels located on the left- and right-hand sides of the water vapor channel, respectively; Twv is the transmittance of the absorption channel; and ρwv*, ρ01* and ρ02* are the apparent reflectances of the absorption channel and two window channels, respectively. The proportional constants, k01 and k02 are calculated as

    where λwv is the wavelength of the absorption channel; λ01 is the wavelength of the left-hand window channel—that is, shorter wavelength; and λ02 is the wavelength of the right-hand window channel—that is, longer wavelength. The water absorption channels of MERSI-II are 905, 936 and 940 nm, and the left- and right-hand window channels are 865 and 1030 nm, respectively. Note that the premise of Eqs. (3) and (4) is that the transmittance of the window channel equals one; otherwise, Twv should be determined based on the transmittance of the window channel (T01) or the weighted sum of the transmittance of two window channels (k01T01 + k02 T02).

    The transmittances of the five MERSI-II channels as a function of total water vapor amount were precalculated using MODTRAN4.3. The transmittances were calculated under six different atmospheric conditions by using the six standard atmospheric models defined in MODTRAN4.3 (Berk et al., 1987), which represent different latitudes worldwide as well as different water vapor levels. Figure 2b shows examples of the transmittances of the five MERSI-II channels as a function of the total water vapor content based on the tropical atmospheric model. The actual water vapor transmittance was extracted from the radiances of the five MERSI-II channels using two- or three-channel ratio techniques according to the surface type. The three-channel ratio technique was used for water vapor derivations over clear land pixels (the reflectance spectrum linearly correlates with the wavelength), while the two-channel ratio technique was used over the ocean with sun glint and cloud pixels [the reflectance spectrum is almost constant (Kaufman and Gao, 1992)]. Subsequently, the combined two-way water vapor content ($ {W}^{\mathrm{*}} $) in the sun–surface–sensor path, called the slant total water vapor content, was derived from the transmittance–water vapor lookup table. The selection of the transmittance–water vapor lookup table was based on the solar zenith angle and surface temperature (see Table 3). The slant total water vapor content, $ {W}^{\mathrm{*}} $, was then converted to the vertical column water vapor content, W, based on the solar and observational geometries using the following equation:

    ConditionAtmospheric model usedConditionAtmospheric model used
    SZA ≤ 20Tropical60 < SZA ≤ 70Subarctic summer
    20 < SZA ≤ 45Midlatitude SummerSZA > 70, T < T0Subarctic winter
    45 < SZA ≤ 60Midlatitude WinterSZA > 70, TT0U.S. standard

    Table 3.  Selection of the transmittance–water vapor lookup table according to the solar zenith angle (SZA) and surface temperature (T).

    where θs and θv are the solar and view zenith angles, respectively.

    As shown in Fig. 2, water vapor has very different absorption coefficients over the three absorption channels. Therefore, under a given atmospheric condition, the derived water vapor values from the three channels can be different. A mean water vapor value (W) is obtained from the following equation:

    where W1, W2 and W3 are the water vapor values derived from the 905, 936 and 940 nm channels, respectively; and f1, f2 and f3 are the corresponding weighting functions, which are calculated based on the sensitivity of the transmittance in each channel (Twv,i) to the water vapor (W): ηi = ΔTwv,i / ΔW. The weighting functions, $ {f}_{i} $, are defined as the normalized values of ηi (Gao and Kaufman, 2003; Hu et al., 2011):

    These normalized weighting functions are numerically computed using the simulated transmittance–water vapor curves. Figure 5 shows the dependence of $ {f}_{i} $ on the total PWV for the three absorption channels of MERSI-II.

    Figure 5.  Normalized weighting factors as a function of the total PWV for the three absorption channels of MERSI-II.

  • In this study, the GPS-derived PWV is used as a reference to compare with the PWV retrieved from MERSI-II NIR channels. Compared with the traditional ground-based water vapor instrument, i.e., radiosonde, GPS has the advantage of a high observation frequency. The temporal resolution of the radiosonde is relatively low, only twice per day. However, with GPS techniques, the PWV has been successfully retrieved at a high temporal resolution (e.g., 10 or 30 min). This helps to increase the opportunity to validate the satellite products and also reduces the impact of the differences between the satellite and ground-based instruments. The GPS computes the atmospheric water vapor column from the phase delays of GPS signals induced by the ionosphere and neutral atmosphere (Bevis et al., 1992). The GPS PWV data were obtained from SuomiNet (http://www.suominet.ucar.edu/data.html). SuomiNet, a university-based, real-time, national GPS network, is funded by the National Science Foundation and costs are shared with collaborating universities. It has been designed for real-time atmospheric remote sensing. The network includes over 100 GPS stations globally, providing near-real time PWV data every 30 min, with a typical accuracy of ~1–2 mm (Tregoning et al., 1998; Ware et al., 2000; Li, 2003; Elgered et al., 2005).

  • Time and space co-located GPS data are used to evaluate the MERSI-II PWV products. The setting of criteria for the spatiotemporal collocation between satellite and ground-based data is important, as it may introduce some errors and influence the comparison results. The temporal difference is usually limited from ±30 min to ±4 h of satellite overpass time and the spatial distance discrepancy is usually limited to a window size from 3 km × 3 km to 0.5° × 0.5° between the satellite pixels and ground-based stations (Liu et al., 2006, 2013, 2015b; Wong et al., 2015; Gong et al., 2018). In this study, the collocation is limited within ±1 h of time coincidence and a distance window of 5 km radius between the GPS location and the center of the MERSI-II retrievals. Thus, the GPS measurements within ±1 h of the MERSI-II overpass time were averaged to compare the results with MERSI-II products. The MERSI-II pixels with a central location of 5 km within the coordinates of the ground-based stations were averaged for further comparison. Note that the MERSI-II pixels used for the comparison should pass through the cloud mask filtering and the fraction of the remaining valid cloud-free pixels must exceed 90%. The MERSI-II L2 Cloud Mask (CLM) products obtained at the same time as the MERSI-II L2 PWV products were used to screen cloudy pixels. In the CLM images, pixels with values of 0, 1, 2 and 3 indicate cloudy, probably cloudy, probably clear, and clear, respectively. Hence, pixels with CLM values of 0, 1 and 2 were excluded from further analysis.

    Four statistical indicators, including the correlation coefficient (R), root-mean-square error (RMSE), mean bias (MB), and relative bias (RB), were used to evaluate the MERSI-II PWV products. The RMSE was used to quantify the differences between paired datasets and is defined as

    where n is the number of data pairs; PWVM represents the MERSI-II PWV data; and PWVR represents the reference PWV—that is, the GPS-derived PWV.

    The MB and RB provide information on overestimation or underestimation of the dataset and are defined as follows:

    The R indicates the strength of the linear relationship between paired datasets and can be calculated using Eq. (13):

    When RB and MB are close to 0 and the RMSE is small, the accuracy of the MERSI-II PWV products is considered to be high. A high R also suggests a good agreement between the two sets of data.

4.   Results and discussion
  • To investigate the PWV retrieval error caused by the radiometric calibration accuracy, four retrieval experiments were performed by setting the water absorption channel as having a positive calibration bias of 5%, 10% and 15%, respectively. The two-channel ratio technique and the transmittance–water vapor lookup table for the tropical model atmosphere were used here to retrieve the water vapor content. The water vapor amounts in the lookup table were used as reference PWV. Based on the principle of the two-channel ratio technique illustrated in Eq. (3), the transmittance at the water absorption channel is directly proportional to the TOA reflectance of this channel, which is proportional to the calibration coefficient. Hence, the transmittance in the lookup table was scaled according to the calibration bias to obtain a new set of transmittance values. Subsequently, a new set of PWV contents was obtained by searching the lookup table. Figures 6ac present comparisons between the newly retrieved PWV amounts and reference PWV values for the three absorption channels. The results in Fig. 6 show that the retrieved PWV values are lower than the reference PWV when the water absorption channel has a positive calibration bias. The retrieved PWV amounts decrease with increasing calibration bias of the water absorption channel—that is, the transmittance increases, which agrees well with the relationship between the transmittance and PWV content shown in Fig. 2. The absolute PWV retrieval error is equivalent to the calibration bias. The PWV retrieval error of the weak absorption channel of 905 nm is the largest among the errors of the three absorption channels under the same calibration bias conditions. The retrieval errors in terms of the RB are −9.1%, −16.7% and −23.1% when the calibration bias is 5%, 10% and 15%, respectively. The PWV retrieval error of the narrow strong absorption channel at 936 nm is the smallest and the RB values are −5.9%, −11.3% and −16.2%, respectively. Those of the channel at 940 nm are slightly larger, at −6.8%, −12.9% and −17.4%, respectively. Because the PWV retrieval error is inversely proportional to the calibration bias, the magnitude of the PWV retrieval error is the same as the value reported here, but the sign is reversed.

    Figure 6.  Scatterplots of the reference water vapor values and those retrieved when the water absorption channel has a positive calibration bias of 5%, 10% and 15%: (a) 905 nm; (b) 936 nm; and (c) 940 nm. The gray dashed line has a 1:1 ratio and the blue line represents the PWV values for the water absorption channel with a positive calibration bias of 5% with respect to the reference PWV. The green and red lines represent the cases with a calibration bias of 10% and 15%, respectively.

    The actual MERSI-II L1B data with two different sets of calibration coefficients (Table 2) were used to further evaluate the influence of the radiometric calibration on the PWV product precision. One of the calibration coefficients represents the pre-launch calibration test and the other represents the on-orbit operational calibration results. Based on the RB in the calibration between the pre-launch and on-orbit operational calibration coefficients (listed in Table 4), the RB of the water vapor transmittance used for the PWV retrieval can be derived through Eqs. (3) and (4). As listed in Table 4, the calculated transmittance values are positively biased; the values for the 905, 936 and 940 nm channels are 6.58%–9.15%, 11.49%–16.35%, and ~7.49%, respectively. Based on the assumption that the calibration of the window channel is correct, the error in the transmittance of the absorption channel is equal to the calibration error in this channel. The simulation analysis between the calibration bias and PWV retrieval accuracy in Fig. 6 indicates that the PWV retrievals of the 905, 936 and 940 nm channels will be underestimated by approximately 9.1%–16.7%, 11.3%–16.2%, and 12.9%, respectively.

    Band No.Center wavelength (nm)RB of the calibration (%)RB of the transmittance (%)
    Two-channel ratioThree-channel ratio
    15*8650.00
    169056.586.589.15
    1793611.4911.4916.35
    18*9400.000.007.49
    191030−9.68
    Note: *For channels 15 and 18, the pre-launch test results were used as the on-orbit operational calibration coefficients. Hence, there is no difference in the calibration results between the pre-launch and on-orbit operational calibration coefficients. For the two-channel ratio technique, the window channel with the shorter wavelength (865 nm) was adopted to calculate the PWV transmittance.

    Table 4.  RB of the calibration and water vapor transmittance for the pre-launch and on-orbit operational calibration coefficients.

    The accuracies of the MERSI-II PWV products were validated using global GPS measurements. The global MERSI-II L1B data were collected during the on-orbit test period from 25–29 May 2018, calibrated using the two different sets of calibration coefficients, and then used to generate the PWV products based on the water vapor retrieval algorithm mentioned in section 3. The left-hand panels in Fig. 7 present scatterplots between the GPS-derived PWV values and those from MERSI-II based on the pre-launch calibration coefficients. The right-hand panels are similar but based on the use of the operational calibration coefficients to calibrate the MERSI-II L1B data. The comparison shows that the PWV retrieval error of channels 905 and 936 nm is large when the pre-launch calibration coefficients are used, with a RB of −37.8% and −30.9%, respectively, and that of the 940 nm channel is relative lower, with a RB of −15.6%. After updating the calibration coefficients, the error of the MERSI-II retrieved PWV values for channels 905, 936 and 940 nm decreased to −20.8%, −12.9% and −8.0%, respectively, and the product accuracy increased by 17.0%, 18.0% and 7.6%, respectively. These values agree well with the above-mentioned analysis of the difference in the accuracy of the PWV retrievals based on the use of pre-launch and on-orbit operational calibration coefficients. The validation results in Fig. 7 also illustrate that, by using the current operational calibration coefficients, good accuracies of the final PWV products—that is, the three-channel weighted ones [Eq. (8)]—are achieved, with an RMSE and RB of 3.42 mm and −13.36%, respectively, which are comparable with values obtained for other global remote sensing products of the same type, such as the NIR PWV products from MODIS. Typical RB and RMSE values in the MODIS NIR PWV products are 5%–10% and less than 3 mm, respectively (Gao and Kaufman, 2003). Larger errors, with RB > 10% and RMSE between 4 and 13 mm, have also been reported for MODIS under humid conditions, such as in summer or in coastal areas (Liu et al., 2013, 2015b; Wong et al., 2015).

    Figure 7.  Scatterplots between the water vapor values measured with GPS and those retrieved from MERSI-II NIR channels for the period 24–29 May 2018 globally: (a–d) pre-launch calibration coefficients used to calibrate MERSI-II L1B data; (e–h) on-orbit operational calibration coefficients used to calibrate MERSI-II Level-1B data. The gray dashed line has a 1:1 ratio.

  • The MERSI-II NIR PWV products have been routinely produced at the National Satellite Meteorological Center, China, since June 2018, several months after the launch. The products can be downloaded from the FengYun Satellite Remote Sensing Data Service Network (http://satellite.cma.gov.cn/portalsite/default.aspx). The operational NIR PWV products include the Level-2 5-min granule product and Level-3 global daily, 10-day, and monthly mean products. The Level-2 products are generated on a pixel-by-pixel basis (i.e., 1 km × 1 km) from standard MERSI-II L1B radiance datasets as well as ancillary data from the L1B geolocation and Level-2 cloud-mask datasets. The Level-1B radiance dataset includes the radiances of five MERSI-II channels centered at 865, 905, 936, 940 and 1030 nm, and the surface temperature based on thermal channels. The Level-1B geolocation Level-2 cloud mask datasets contain ancillary data including solar zenith angle, view zenith angle, solar azimuth angle, view azimuth angle, land–sea flag, and cloud mask. The outputs of the Level-2 PWV product include the total weighted column water vapor content and that independently derived from one of the three absorption channels and an associated quality assurance parameter, which indicates whether the inversion algorithm has a two-channel or three-channel ratio and whether a pixel is clear or cloudy. In addition to the Level-2 NIR water vapor product, the Level-3 daily NIR PWV products for clear and cloudy areas are produced based on a mosaic and projection of the 5-min granule global products and have a spatial resolution of 0.05° × 0.05°. The ten-day and monthly products are then produced based on the average of the daily products; they also have a spatial resolution of 0.05° × 0.05°.

    Figures 810 show a few examples of Level-2 5-min granule water vapor images and Level-3 daily and monthly mean water vapor images. Figure 8b presents an example of a Level-2 PWV image over the Arabian Peninsula obtained at 1000 UTC 13 September 2019. The corresponding true color image composite processed from the MERSI-II L1B radiance of channels centered at 650 nm (red), 550 nm (green) and 470 nm (blue) is shown in Fig. 8a. The Arabian Peninsula is located at 13°N–20°N and is a flat plateau with an average altitude of 1200–2500 m. This region is characterized by a tropical desert climate with little precipitation. The hot and dry climate results in a large desert within this region, accounting for approximately one-third of the total area (Fig. 8a, yellow–brown colors). The PWV image in Fig. 8b illustrates that only the parts close to water bodies have relatively higher moisture contents, with water vapor contents in the range of 20–40 mm, while the water vapor concentrations across the rest of the area are small (mostly < 20 mm), which is consistent with the dry climate characteristics. Further, the variation of the PWV over the inland area is weak because of the flat plateau.

    Figure 8.  (a) Example of a MERSI-II Level-2 PWV image over the Arabian Peninsula obtained at 1000 UTC 13 September 2019 and (b) the corresponding true color image composite processed from the MERSI-II Level-1B radiance of channels centered at 650 nm (red), 550 nm (green) and 470 nm (blue). There are no PWV retrievals over the water bodies, except for the sun-glint region of the ocean.

    Figure 10.  Monthly averaged global Level-3 water vapor images derived from FY3D MERSI-II NIR channel observations in (a) July 2019, (b) October 2019, and (c) December 2019.

    Figure 9.  Daily global Level-3 water vapor images derived from FY3D MERSI-II NIR channels on 2 July 2019.

    Figure 9 is an example of a MERSI-II Level-3 daily global PWV image obtained on 2 July 2019. There are no PWV retrievals over the ocean, except for the sun-glint area (oval region in Fig. 9). Regions with solar zenith angles above 72° also has no retrievals, as the observation time is close to night and the energy at the satellite’s entrance is weak. Figure 9 shows that the daily PWV product contains fine spatial variation patterns of the water vapor and thus can be used for the analysis of the spatial distribution of the PWV over a large area. A high PWV content is typically observed near the equator. High PWV values (> 35 mm) are also detected in the land area, such as in the Amazon rainforest in South America, South Asia, southeastern China, Southeast Asian islands, and central and southern Africa. Furthermore, sub-high water vapor concentrations (> 25 mm) can be detected in northeastern China, central Europe, southern Russia, and the southeastern United States. Because July corresponds to winter in the Southern Hemisphere, the water vapor contents over the southern parts of South America and Africa, as well as Australia, are small (< 20 mm).

    Figure 10 presents examples of Level-3 monthly averaged PWV images for the entire globe for the months of July, October and December in 2019. The PWV contents in eastern China increase with decreasing latitudes. The interannual variation of the PWV distribution is significant in midlatitude areas. From July to December—that is, from summer to winter in the Northern Hemisphere—a decrease in the PWV content can be observed in the southeastern United States, southern parts of India, and southeastern China. In contrast, a slight increase in the PWV content can be observed in the midlatitude regions of the Southern Hemisphere because the seasonal changes are opposite to those in the Northern Hemisphere. In the tropical and high-latitude areas, the interannual variation of the PWV distribution is weak. The analysis of the water vapor images in Figs. 10ac shows that the Level-3 monthly mean PWV images can be used to study the seasonal variations of water vapor on a global scale.

  • To validate the accuracy of the NIR PWV products, we compared the MERSI-II NIR water vapor values with those measured by the ground-based instrument, i.e., GPS, from January to December 2019. The GPS sites located in North America were used for comparison because data updates for other global sites have stopped since October 2018 to this June. Figure 11a shows a scatterplot of the water vapor values measured with the GPS and those measured with MERSI-II on clear days. A high correlation coefficient (R2 = 0.96) was obtained, suggesting good agreement between the two sets of data. The RMSE, MB and RB were 3.87 mm, −2.31 mm and −10.87%, respectively, indicating a relatively small systematic bias. Figure 11b presents a histogram of the PWV retrieval error (RB) in different channels. In general, the peak value of RB in all three absorption channels ranges from −9% to −16%, with 905 nm a sight larger than the other two channels. Hence, the RB of the final PWV product can be expected to be controlled at less than −10%, through improving the PWV retrieval accuracy at 905 nm. Overall, however, the accuracy of the current NIR PWV products of MERSI-II has significantly improved compared with the previous MERSI. In previous studies, a negative RB of 30%–40% was reported for the FY-3A/MERSI NIR PWV products (Hu et al., 2011; Gong et al., 2018).

    Figure 11.  (a) Scatterplot of the water vapor values measured with the GPS and MERSI-II NIR PWV values measured on clear days from January to December 2019 over North America. (b) Histogram of the difference between the GPS-measured and MERSI-II-retrieved PWVs using different channels.

    The monthly mean RB and RMSE of the MERSI-II NIR PWV products on clear days are shown in Fig. 12. The error variations show an apparent seasonal pattern. The largest error appears in summertime, with an RB and RMSE of −14.3% and 5.5 mm, respectively. The smallest error appears in winter, with an RB and RMSE of around −3.0% and 1.8 mm, respectively. Overall, during the one-year period after the on-orbit test, the RB and RMSE of MERSI-II NIR PWV products were better than −15% and 5.5 mm, respectively, which is consistent with the accuracy obtained during the on-orbit test, suggesting a good calibration accuracy and high radiometric stability of the five corresponding channels of the MERSI-II instrument used for NIR PWV remote sensing.

    Figure 12.  (a) Monthly mean RB and (b) RMSE of the MERSI-II NIR PWV products compared with the GPS measurements from January to December 2019 over North America.

5.   Conclusions
  • MERSI-II is an advanced instrument onboard the latest Chinese FY satellite, FY-3D, which was launched in November 2017. Compared with the previous version of the instrument, MERSI, many updates have been made to MERSI-II, leading to significant improvements in its performance and observation capabilities. The main focus of this paper was the instrument’s ability in providing global PWV observations based on TOA radiance information from NIR channels. Hence, the major improvements related to the NIR PWV retrieval of MERSI-II have been discussed in detail. Furthermore, the water vapor retrieval algorithm based on the use of MERSI-II NIR channels, its current status, and samples of MERSI-II’s PWV products, are presented. Finally, the accuracy of MERSI-II’s PWV products was evaluated by comparing them with the PWV derived from ground-based GPS over a long time period.

    Regarding the instrument improvements related to NIR PWV retrieval, the settings of the water vapor absorption channels are more reasonable than those used in the previous version, MERSI. In MERSI, two weak absorption channels at 905 and 980 nm were adopted. The simulated transmittances show that these two channels are both more sensitive to changes in the concentration of atmospheric water vapor under dry conditions. To avoid redundant information, only one weak absorption channel at 905 nm was kept in MERSI-II and the 980 nm channel was replaced with a wide and strong absorption channel at 940 nm. The sensitivity to changes in the water vapor content of this channel is between that of the other two channels, providing additional information for the water vapor retrieval under intermediate water vapor conditions. Accordingly, MERSI-II provides more information under different conditions (dry, intermediate, and humid).

    The radiometric calibration accuracy affects the precision of the PWV products retrieved using NIR channels. The absolute value of the PWV retrieval error is equivalent to the calibration bias; the error of the 905 nm channel is slightly larger than that of the other two water absorption channels. A WD-IVC calibration method was applied for on-orbit operational calibration updates in MERSI-II, leading to a decrease in the calibration uncertainty based on the use of different types of automatic on-orbit VC methods. Based on the use of the WD-IVC method, the long-term calibration accuracy of the five channels used for the PWV retrieval has been maintained within ±5% since MERSI-II’s launch.

    Benefitting from MERSI-II’s better channel setting in the NIR spectral region as well as sound and stable calibration accuracy, the quality of the NIR PWV products of MERSI-II has significantly improved compared with that of the previous MERSI. The RB and RMSE of the PWV values were estimated to be better than −15% and 5.5 mm, respectively, one year after the launch. The granule Level-2 “pixel-based” NIR PWV product and the daily, 10-day and monthly Level-3 NIR PWV products (global 0.05° × 0.05° latitude–longitude grid) have been routinely produced and archived at the National Satellite Meteorological Center computing facility since June 2018. These products are useful for the study of seasonal and annual variations of water vapor content on regional and global scales.

    The small systematic underestimation of the current NIR PWV products will be further investigated in the next step. For example, the precision of the transmittance calculation by MODTRAN4.3, the uncertainty in the surface spectra reflectance, and the influence of high aerosol loadings (e.g., dust, haze) as well as thin cirrus clouds on the PWV retrieval should be studied in the future.

    Acknowledgements. This research was funded by the National Key R&D Program of China (Grant Nos. 2018YFB0504900, 2018YFB0504901, and 2018YFB0504802) and the National Natural Science Foundation of China (Grant Nos. 41871249 and 41675036). The authors thank SuomiNet for providing the GPS data, as well as the FengYun Satellite Remote Sensing Data Service Network for providing the FY-3D/MERSI-II data.

Reference

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return