• Alquaied, F., and W. L. Jones, 2017: An emissive antenna correction for the Tropical Rainfall Measuring Mission Microwave Imager. 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 2104–2107, https://doi.org/10.1109/IGARSS.2017.8127398.

    • Crossref
    • Export Citation
  • Alquaied, F., R. Chen, and W. L. Jones, 2017: Hot load temperature correction for the Tropical Rainfall Measuring Mission Microwave Imager (TMI). 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 2108–2111, https://doi.org/10.1109/IGARSS.2017.8127399.

    • Crossref
    • Export Citation
  • Berg, W., and Coauthors, 2016: Intercalibration of the GPM radiometer constellation. J. Atmos. Oceanic Technol., 33, 26392654, https://doi.org/10.1175/JTECH-D-16-0100.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bilanow, S., and S. Slojkowski, 2006: TRMM on-orbit performance reassessed after control change. Extended Abstracts, 19th Int. Symp. on Space Flight Dynamics, Kanazawa City, Japan, JAXA, ISTS 2006-d-35, 3 pp., https://pps.gsfc.nasa.gov/Documents/Bilanow_TRMM_Pointing_Control_ISTS_2006-d-35_a.pdf.

  • Biswas, S. K., K. Gopalan, W. L. Jones, and S. Bilanow, 2010: Correction of time-varying radiometric errors in TRMM Microwave Imager calibrated brightness temperature products. IEEE Geosci. Remote Sens. Lett., 7, 851855, https://doi.org/10.1109/LGRS.2010.2050135.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Draper, D. W., D. A. Newell, D. S. McKague, and J. R. Piepmeier, 2015: Assessing calibration stability using the Global Precipitation Measurement (GPM) Microwave Imager (GMI) noise diodes. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 8, 42394247, https://doi.org/10.1109/JSTARS.2015.2406661.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ji, Y., S. Bilanow, and E. Stocker, 2008: NASA TRMM TMI level 1B algorithm, version 1. TRMM Science and Data Information System Algorithm Theoretical Basis Doc., 40 pp., https://pps.gsfc.nasa.gov/Documents/L1B11ATBD_V1.pdf.

  • Kroodsma, R., S. Bilanow, Y. Ji, and D. McKague, 2017: TRMM Microwave Imager (TMI) updates for final data version release. 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 1216–1219.

  • Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson, 1998: The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Oceanic Technol., 15, 809817, https://doi.org/10.1175/1520-0426(1998)015<0809:TTRMMT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kummerow, C., J. Simpson, O. Thiele, W. Barnes, A. T. C. Chang, E. Stocker, and R. F. Adler, 2000: The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit. J. Appl. Meteor., 39, 19651982, https://doi.org/10.1175/1520-0450(2001)040<1965:TSOTTR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Skofronick-Jackson, G., and Coauthors, 2017: The Global Precipitation Measurement (GPM) mission for science and society. Bull. Amer. Meteor. Soc., 98, 16791695, https://doi.org/10.1175/BAMS-D-15-00306.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stocker, E., 2016: File naming convention for precipitation products for the Global Precipitation Measurement (GPM) mission. NASA GSFC PPS_610.2_P550, Version 1.4.6, 22 pp., https://pps.gsfc.nasa.gov/Documents/FileNamingConventionForPrecipitationProductsForGPMMission.pdf.

  • Takahashi, N., and T. Iguchi, 2004: Estimation and correction of beam mismatch of the precipitation radar after an orbit boost of the Tropical Rainfall Measuring Mission satellite. IEEE Trans. Geosci. Remote Sens., 42, 23622369, https://doi.org/10.1109/TGRS.2004.837334.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 2015: A 17-year climate record of environmental parameters derived from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager. J. Climate, 28, 68826902, https://doi.org/10.1175/JCLI-D-15-0155.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., and D. Draper, 2016: On-orbit absolute calibration of the Global Precipitation Measurement Microwave Imager. J. Atmos. Oceanic Technol., 33, 13931412, https://doi.org/10.1175/JTECH-D-15-0212.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., P. Ashcroft, and C. Gentemann, 2001: Post-launch calibration of the TRMM Microwave Imager. IEEE Trans. Geosci. Remote Sens., 39, 415422, https://doi.org/10.1109/36.905249.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    Sample (left) preboost and (right) postboost orbit: (top) pitch, (middle) roll, and (bottom) yaw, showing onboard (black), version 7 (blue), and version 8 attitudes (green).

  • View in gallery

    Minimum (blue), maximum (red), and mean (green) values for (top) pitch, (middle) roll, and (bottom) yaw for each orbit over the TRMM mission span.

  • View in gallery

    Block diagram of the two-point calibration radiometer.

  • View in gallery

    Orientation of TMI antenna beams during the deep-space calibration roll maneuver. The black arrow is the main beam, the red one the reflector spill-over beam, and the green one the cold sky mirror beam.

  • View in gallery

    (a) Antenna temperature during four phases of the DSM, and (b) the expanded phases 3 and 4.

  • View in gallery

    Flowchart of the procedure used to estimate TMI MR physical temperature during normal operation.

  • View in gallery

    The physical temperature of the TMI’s MR for a solar beta angle = 43.5°.

  • View in gallery

    The physical temperature of the TMI MR.

  • View in gallery

    TMI calibration gain of TMI 18-GHz H channel. (top) Version 7 single, and (bottom) version 8 multiscan calibration with hot load and cold load corrections.

  • View in gallery

    TMI 18-GHz H channel Ta differences between versions 8 and 7.

  • View in gallery

    RFI flag for 10-GHz Earth samples.

  • View in gallery

    RFI flag for 19-GHz Earth samples for the same area shown in Fig. 11.

  • View in gallery

    The Tb of 10 GHz for the same orbit as in Figs. 11 and 12.

  • View in gallery

    TMI cold load count of TMI 18-GHz H channel. (top) Version 7 single-scan calibration, and (bottom) version 8 multiscan calibration with RFI correction.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 180 164 7
PDF Downloads 149 138 8

TRMM Version 8 Reprocessing Improvements and Incorporation into the GPM Data Suite

View More View Less
  • 1 NASA GSFC, Greenbelt, Maryland
  • | 2 Central Florida Remote Sensing Laboratory, University of Central Florida, Orlando, Florida, and King Abdulaziz City for Science and Technology, Riyadh, Saudi Arabia
  • | 3 NASA GSFC, and Stinger Ghaffarian Technologies, Inc., Greenbelt, Maryland
  • | 4 Central Florida Remote Sensing Laboratory, University of Central Florida, Orlando, Florida
© Get Permissions
Full access

Abstract

The National Aeronautics and Space Administration (NASA) has always included data reprocessing as a major component of every science mission. A final reprocessing is typically a part of mission closeout (known as phase F). The Tropical Rainfall Measuring Mission (TRMM) is currently in phase F, and NASA is preparing for the last reprocessing of all the TRMM precipitation data as part of the closeout. This reprocessing includes improvements in calibration of both the TRMM Microwave Imager (TMI) and the TRMM Precipitation Radar (PR). An initial step in the version 8 reprocessing is the improvement of geolocation. The PR calibration is being updated by the Japan Aerospace Exploration Agency (JAXA) using data collected as part of the calibration of the Dual-Frequency Precipitation Radar (DPR) on the Global Precipitation Measurement (GPM) Core Observatory. JAXA undertook a major effort to ensure TRMM PR and GPM Ku-band calibration is consistent.

A major component of the TRMM version 8 reprocessing is to create consistent retrievals with the GPM version 05 (V05) retrievals. To this end, the TRMM version 8 reprocessing uses retrieval algorithms based on the GPM V05 algorithms. This approach ensures consistent retrievals from December 1997 (the beginning of TRMM) through the current ongoing GPM retrievals. An outcome of this reprocessing is the incorporation of TRMM data products into the GPM data suite. Incorporation also means that GPM file naming conventions and reprocessed TRMM data carry the V05 data product version. This paper describes the TRMM version 8 reprocessing, focusing on the improvements in TMI level 1 products.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Erich Stocker, erich.f.stocker@nasa.gov

This article is included in the Global Precipitation Measurement (GPM) special collection.

Abstract

The National Aeronautics and Space Administration (NASA) has always included data reprocessing as a major component of every science mission. A final reprocessing is typically a part of mission closeout (known as phase F). The Tropical Rainfall Measuring Mission (TRMM) is currently in phase F, and NASA is preparing for the last reprocessing of all the TRMM precipitation data as part of the closeout. This reprocessing includes improvements in calibration of both the TRMM Microwave Imager (TMI) and the TRMM Precipitation Radar (PR). An initial step in the version 8 reprocessing is the improvement of geolocation. The PR calibration is being updated by the Japan Aerospace Exploration Agency (JAXA) using data collected as part of the calibration of the Dual-Frequency Precipitation Radar (DPR) on the Global Precipitation Measurement (GPM) Core Observatory. JAXA undertook a major effort to ensure TRMM PR and GPM Ku-band calibration is consistent.

A major component of the TRMM version 8 reprocessing is to create consistent retrievals with the GPM version 05 (V05) retrievals. To this end, the TRMM version 8 reprocessing uses retrieval algorithms based on the GPM V05 algorithms. This approach ensures consistent retrievals from December 1997 (the beginning of TRMM) through the current ongoing GPM retrievals. An outcome of this reprocessing is the incorporation of TRMM data products into the GPM data suite. Incorporation also means that GPM file naming conventions and reprocessed TRMM data carry the V05 data product version. This paper describes the TRMM version 8 reprocessing, focusing on the improvements in TMI level 1 products.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Erich Stocker, erich.f.stocker@nasa.gov

This article is included in the Global Precipitation Measurement (GPM) special collection.

1. Introduction

The National Aeronautics and Space Administration (NASA), as a research-driven engineering and science innovator, has always included the reprocessing of mission data products as part of all mission phases. Within NASA missions, reprocessing allows for all the data collected during a mission to be reprocessed using the latest algorithms, ensuring that consistent data products are generated for the entire mission.

During the course of its more than 17 yr of operations, the Tropical Rainfall Measuring Mission (TRMM) (Kummerow et al. 2000) has had seven data product versions. The first three were within the instrument calibration period and were not distributed, but each one processed the mission data from the date of instrument turn-on. The first public version was version 4. This operational version provided a consistently processed mission dataset. The last reprocessing during the TRMM operational phase was version 7 in July 2011. Version 7 data exist from the beginning of the mission to the end of the mission in April 2015.

With the end of the TRMM satellite, the mission entered the closeout phase (known in NASA parlance as phase F) that includes a final reprocessing of the TRMM data. This version is TRMM, version 8, but with its incorporation into the Global Precipitation Measurement (GPM) data suite, all of these reprocessed TRMM products are identified with GPM data product, version 05 (V05). As part of the incorporation into the GPM data suite, TRMM data products are renamed using GPM conventions.

In February 2014, the GPM (Skofronick-Jackson et al. 2017) Core Observatory was launched by the Japan Aerospace Exploration Agency (JAXA) from the Tanegashima Space Center, Japan. From its very conception as a mission, GPM always planned to create a consistent dataset that extended back to the beginning of the TRMM mission and to apply the latest GPM algorithms to TRMM-era data. The 1-yr overlap of the TRMM and GPM Core satellites was key to establishing consistent reflectivities, brightness temperatures (Tb), and precipitation retrievals between the two missions.

The TRMM Microwave Imager (TMI), a conical scanning microwave radiometer, is composed of nine channels with five different frequencies, namely, 10.65, 19.35, 21.30, 37.0, and 85.5 GHz. All frequencies were dual polarized (pol)—vertical (V) and horizontal (H)—except for 21.3 GHz, which was V-pol only (Kummerow et al. 1998). The GPM Microwave Imager is a conically scanning microwave radiometer on board the GPM Core Observatory. The Core Observatory flies in a 407-km circular orbit with an inclination angle of 65°. The GMI has 13 channels at frequencies of 10.65, 18.7, 23.8, 36.64, 89, 166, and 183.31 GHz (Skofronick-Jackson et al. 2017).

The intercalibration of TMI on TRMM to the GPM Microwave Imager (GMI) on the GPM Core Observatory serves to establish TRMM as the calibration transfer standard for earlier satellites that did not overlap the GPM era. This intercalibration leads to consistent brightness temperatures among the TRMM and GPM constellation microwave radiometers. These are used in producing consistent global precipitation products; they also facilitate merging data from the many microwave radiometers into a global product.

A major effort in the TRMM version 8 reprocessing is the improvement of TMI by addressing a number of issues, such as geolocation, emissive antenna corrections, multiscan calibration, and radio frequency interference (RFI). The details of the most important of these corrections are provided in the following sections.

At the end of the TRMM operational period, deep-space calibration maneuvers were performed that were instrumental in developing the TMI emissive antenna corrections explained in section 4. These maneuvers were similar to ones performed early in the mission. So, they also verified that TMI remained very consistent over its more than 17 years of operation. This means it is useful as a calibrator for other microwave radiometers that flew during the TRMM era.

The TRMM Precipitation Radar, which consists of a Ku-band phased array system, was subject to postlaunch calibration changes in the level 1 processing for the GPM V05 reprocessing. This resulted in an overall increase of approximately +1.1 dB, directly impacting surface cross-sectional and measured reflectivity. The calibration change was the result of JAXA active radar calibrations performed postlaunch and comparisons with the GPM Dual-Frequency Precipitation Radar Ku-band instrument during an overlap period in 2014.

The altitude increase of TRMM from 305 to 402 km in 2001 resulted in a beam mismatch between transmitted and received echoes caused by the fixed beam switching of PR and resulted in a low echo bias on one side of the across-track scan, which in turn yielded an asymmetry in precipitation detection. This bias was partially mitigated in TRMM version 7 (Takahashi and Iguchi 2004), but a further review using a new method resulted in improved precipitation detection and bias corrections (K. Kanemaru 2017, personal communication). The 17+-yr record of TRMM PR showed a time-dependent drift of approximately 0.2 dB in the surface cross section over ocean, which has also been mitigated in GPM V05 (Takahashi and Iguchi 2004). Further details on the calibration changes for GPM V05 processing of PR are outside the scope of this work and will be detailed in another publication.

This paper provides information about the work undertaken to incorporate the entire TRMM data record into the GPM data suite. The TRMM version 8 reprocessing and level 1 data product improvement are major parts of this integration effort. While all TRMM algorithms are being migrated to GPM-based algorithm code, this paper concentrates primarily on the work to improve the TMI level 1 data products.

2. Format and file name changes

a. Format changes

Upon completion of TRMM version 8 reprocessing, the data products will become part of the GPM data suite. To this end, the TRMM data product formats had to be migrated to formats consistent with the ones used in GPM V05. The inclusion of TRMM version 8 reprocessed products into the GPM V05 data suite extends the GPM data suite back to 1998.

The most fundamental format change is the change from Hierarchical Data Format version 4 (HDF4) to HDF version 5 (HDF5). All GPM data are stored in HDF5. With this TRMM reprocessing, all TRMM data also are stored in HDF5. HDF4 will no longer be supported.

The TRMM Visible and Infrared Scanner (VIRS) has only two products: A level 1A product that contains the instrument counts in packed format and the level 1B radiance product that has geolocated instrument field-of-view (FOV) radiances for each of its channels. During this final TRMM mission reprocessing, geolocation occurs at level 1A. This change means that the instrument counts are geolocated. Also, the level 1A product is in HDF5 instead of in binary format. VIRS level 1B radiance products have the same content as the TRMM version 7 products but are in HDF5 rather than HDF4.

The TRMM Precipitation Radar (PR) products are undergoing substantial changes. These are not described in detail in this paper, but they will be described in other papers by JAXA scientists and engineers. The radar level 1B product has content almost identical to the GPM Ku-band radar level 1B. The three separate TRMM PR level 2 products (sigma zero, precipitation types, and precipitation retrievals) are combined into a single level 2 product using the GPM V05 Ku-band radar format. In addition, just as in GPM V05, TRMM reprocessed version 8 products include the reflectivities within the single level 2 product rather than as a separate product as was previously done in TRMM. These changes bring the TRMM radar products into conformance with GPM V05.

TRMM TMI also has substantial content changes for version 8. The reprocessed TRMM TMI level 1A product is not packed binary as in TRMM version 7, but it has geolocated counts similar to the GMI level 1A product and is stored in HDF5 format. From the beginning of TRMM, the level 1 brightness temperature products have had only two swaths: One was used to record information about all the low-frequency channels and the second was for recording 85-GHz high-frequency information. As part of the version 8 reprocessing, TMI level 1 brightness temperature products added an additional swath for the 10-GHz channels. The addition of the 10-GHz swath was deemed necessary to account for updated geolocation information. The 10-GHz swath also provides information for two incident angles to deal with the detected differences in the horizontal and vertical channels. All the other low-resolution channels use the format of the second swath. The 85-GHz high-resolution channels comprise the third swath. TMI also adds a 1C product containing intercalibrated brightness temperatures (Berg et al. 2016) using GPM GMI as the calibration reference standard (Wentz and Draper 2016). The TMI 1C product has a format consistent with other GPM microwave radiometers.

The format of the TMI Goddard profiling algorithm (GPROF) precipitation retrievals is identical to all the other GPROF products within the GPM V05 data suite. Users should be aware, however, that it is substantially different from the TRMM version 7 GPROF product.

b. File name changes

Prior to version 8 reprocessing, TRMM always used a unique four-character alphanumeric datatype identifier to uniquely identify each of its products. This identifier is composed of four parts. The key to this identifier is as follows:

  • Number to indicate the processing level.

  • Letter to indicate whether data are from a single sensor or multiple sensors.

  • Number to indicate which instrument (0–VIRS, 1–TMI, 2–PR, 3–multiple).

  • Number to indicate which product for that particular instrument.

For example, the datatype 2A23 is decoded as follows:

  • A 2 denotes level 2 processing (geophysical parameter).

  • An A denotes data from a single instrument (level 1 products are the exception, as A indicates a count product and B a radiance–Tb product).

  • A 2 denotes data comes from PR.

  • A 3 denotes the third product made with PR data (i.e., rain types).

With TRMM version 8 reprocessing, this TRMM datatype identification (ID) concept disappears. TRMM products are named using the GPM file naming convention (Stocker 2016). In this approach the satellite, instrument, and algorithms are specifically listed in the name rather than implied by a code as in the TRMM approach. The TRMM version 7 TMI level 1 brightness temperature product named 1B11.20020203.24079.7.HDF becomes 1B.TRMM.TMI.TB2017.20020203-S200611-E213834.24079.V05A.HDF5 as part of the TRMM version 8 reprocessing. The 1B.TRMM.TMI designators are fixed regardless of the reprocessing version, but the algorithm ID (TB2017) and the data product version (V05A) are changeable during reprocessing.

Last, as part of the file name, the data product version indicates the GPM data version (V05). The adoption of the GPM naming convention in TRMM version 8 reprocessing is necessary for integration into GPM. Now every time GPM goes through a reprocessing cycle, TRMM satellite data will be included.

3. Geolocation improvements

Geolocation data provide the geodetic latitude and longitude of each image pixel. They are computed along with associated ancillary data, such as zenith and azimuth angles of the satellite and sun from each location, based on the spacecraft orbit, attitude, and science instrument view-direction modeling Improvements to the TRMM geolocation are discussed in this section in four categories: ephemeris, attitude, sensor model, and computational improvements. The first three categories are the key inputs for the computations.

a. Ephemeris

Definitive ephemeris files, tables of the spacecraft position versus time, were provided daily by the NASA Goddard Space Flight Center (GSFC) Flight Dynamics Facility (FDF) throughout the TRMM mission. These files used the daily tracking data to provide a best estimate of the satellite path each day in inertial true-of-date coordinates, and they provided in the header data the Greenwich hour angle (GHA) of Earth’s rotation at the start of each day.

Some errors in these ephemeris GHAs were identified in 2013, and a review of the data identified problems in the configuration of the latest Earth orientation parameter (EOP) files at the TRMM Mission Operations Center (MOC). There were also some issues around the times of leap seconds. Worst-case errors amounted to about 450 m—the effect of 1 s of Earth rotation. To support reprocessing, the MOC flight dynamics support personnel regenerated the ephemeris files in Earth rotating coordinates using the latest EOP files from the U.S. Naval Observatory. These corrected ephemeris files form the basis for TRMM version 8 reprocessing.

A measure of the ephemeris accuracy is given by their overlap difference each day. A review by the FDF indicated maximum differences typically less than 0.1 km, occasionally up to 0.5 km, and rarely larger than that, occurring only a few dozen times in the mission. Worst-case errors in accuracy occurred during periods of solar activity that change Earth’s outer atmosphere, adding extra drag on the spacecraft and thereby affecting the orbit. Errors are also greater around the time of orbit-adjust maneuvers.

b. Attitude

The spacecraft attitude describes its orientation in space, expressed as pitch, roll, and yaw for this discussion. Similar to airplane terminology, positive pitch moves the nose up, positive roll moves the left wing up, and positive yaw moves the nose to the right. More specifically, the yaw–pitch–roll are defined as a 3–2–1 Euler sequence in the rotations about spacecraft ZYX axes, respectively starting from a geodetic reference frame. TRMM spent about half the time pointing forward in the flight path (yaw = 0°) and the other half pointing backward (yaw = 180°), as a result of thermal constraints to keep the sun off the +Y side of the spacecraft. The 180° yaw turns occurred every 2–4 weeks as the orbit precession moved the sun above and below the orbit plane.

The onboard Attitude Control System (ACS) was required to provide attitude knowledge to within 0.2° and control to within 0.4° in each axis. These targets were generally met with occasional exceptions, discussed below. The ACS initially used Earth horizon sensors for pitch and roll control, and yaw was updated twice during each orbit using sun sensor data and propagated using Z-axis gyroscope (gyro) data. After the orbit boost in August 2001, the horizon sensors did not work at the higher altitude, and a contingency mode control made use of sun sensor and magnetometer data, along with gyro data, on all three axes incorporated in a Kalman filter. The contingency mode initially did not meet the original requirements, but it managed to meet them well after some postboost tuning of the ACS and enhanced ground support for the onboard ephemeris (Bilanow and Slojkowski 2006). Version 7 adjustments to the attitudes incorporated into ground processing were designed to keep the attitude knowledge errors within the original requirements for the postboost period reprocessing.

For version 8 the spacecraft attitude was recomputed using new ground-definitive attitude computation software that uses spacecraft gyro and sun sensor data along with measurements of the spacecraft roll derived using the PR science data. This approach required a substantial analysis effort and development of a detailed estimate of the gradually changing gyro biases. Extensive efforts were applied to quality assure the solutions from errors caused by miscellaneous data anomalies during the mission.

Features of the onboard attitude, version 7 attitude, and version 8 attitude histories are illustrated in Fig. 1, showing pitch, roll, and yaw for a sample preboost orbit (left) and postboost orbit (right). In certain cases the version 7 attitude is exactly the same as the original onboard attitude. The onboard-reported attitude (black line) stays near zero, since it is used in the ACS control loop and active control targets zero pitch–roll–yaw. Spikes in the onboard attitude occur in response to sudden changes in onboard-estimated attitude, and these are tracked as real shifts in the attitude in the version 8 results. The version 8 attitudes (green lines) illustrate the significant improvements in tracking the true spacecraft attitude time history relative to the onboard and version 7 adjusted attitudes (black and blue lines, respectively).

Fig. 1.
Fig. 1.

Sample (left) preboost and (right) postboost orbit: (top) pitch, (middle) roll, and (bottom) yaw, showing onboard (black), version 7 (blue), and version 8 attitudes (green).

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

In the preboost period, shown on the left, the larger attitude errors in pitch and roll, up to 0.4°, are due to sun interference in the horizon sensor. These types of larger errors typically occurred in partial orbits for a few dozen orbits in a row a few times each month. The errors occurred when the solar beta angle (sun elevation above the orbit plane) was around 45°. Besides these occasional error spikes, the pitch and roll were susceptible to seasonal horizon radiance errors in the tenth-of-a-degree range. Yaw in the preboost period typically showed jumps of 0.1°–0.2° twice each orbit as the yaw was updated using the sun sensor data. The version 7 corrections also took out a sine wave–like component of the errors in roll and yaw because orbit period corrections were picked up with the method applied to correct postboost errors as noted below. The real spacecraft motions as a result of errors and discontinuities in the onboard-calculated attitude are now tracked by using the gyro data with the version 8 ground reprocessing.

In the postboost period sample shown in the right panels of Fig. 1, errors in roll and yaw tended to occur as orbit period sine waves. These are the types of errors that result from gyro propagation from an imperfect initial attitude. The onboard accuracy was limited by systematic magnetic field model inputs and magnetometer uncertainties. For version 7 reprocessing, these sine wave–like errors for roll and yaw were taken out for the postboost period up until 15 January 2010, using data from the PR science products available prior to that reprocessing start (Ji et al. 2008). With the mission now complete, all the available PR data are now used in the new gyro-propagated attitude estimates.

Onboard offsets in pitch pointing in the postboost period were mainly driven by uncertainties in the uplinked onboard ephemeris (used instead of the horizon sensor for Earth direction estimate by the ACS). Pitch was also affected by coupling with roll/yaw errors. Using the new version 8 processing, pitch is mainly obtained from the sun sensor and is tracked by the gyro data, so we now track how the pitch actually changes at each onboard ephemeris uplink, as illustrated about 32 min into the postboost orbit in Fig. 1. Version 7 postboost pitch errors that were typically in the one- or two-tenths-of-a-degree range are now tracked well.

The version 8 attitudes generally improve the accuracy from the tenths-of-a-degree range to the hundredths-of-a-degree range. Errors in the attitude tracking are particularly improved during various anomalous periods. One anomaly occurred 12–14 November 2013 when pitch errors as large as 1° occurred as a result of a failed spacecraft command schedule upload that caused an erroneous onboard ephemeris to be used. The version 8 attitudes now also track the attitudes accurately during periods of special calibration maneuvers, such as the deep-space calibration.

A general picture of the typical variation of the spacecraft attitude in pitch, roll, and yaw with the new processing is given in Fig. 2. This figure shows the minimum, maximum, and mean for each orbit over the TRMM mission lifetime. The different ranges for the spacecraft attitude variation are shown following the orbit boost in August 2001. Some secular drift in the mean attitude in pitch and roll in the several-hundredths-of-a-degree range is shown in the preboost period. Some differences in the mean attitude for the two yaw orientations are also illustrated, especially in roll, where green shows the zero yaw orientations.

Fig. 2.
Fig. 2.

Minimum (blue), maximum (red), and mean (green) values for (top) pitch, (middle) roll, and (bottom) yaw for each orbit over the TRMM mission span.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

One feature near the end of the mission deserves special mention; less variation in the reported roll and yaw are shown in Fig. 2 during the orbit descent phase at the end of 2014 and into 2015. This feature is because the PR instrument science data were not available during that time span for estimated true roll. The current processing reverts effectively to assuming the near-zero onboard roll/yaw estimate in the absence of better data, so the minimum and maximum values are quite small. In reality the true variation would have been typical of that tracked with the benefit of PR science data before and after this span—10 August 2014–12 February 2015. So, the remaining attitude errors in roll and yaw could be expected to be about 0.05°–0.2° during this span. The PR instrument was turned back on after the spacecraft reached about 355-km altitude, and our expected attitude accuracy returns then to the hundredths-of-a-degree range. The PR instrument was kept on until 1 April 2015. Reduced accuracy is expected again for the last week of TMI data collection, up to 8 April 2015, and also for various orbit granules throughout the mission where PR data were not available for various reasons.

A measure of the attitude solution accuracy is given by the overlap difference between each orbit; these differences are generally less than 0.02°. Exceptions are noted for orbits that included yaw turns because gyro propagation through these maneuvers is less accurate. For these orbits, consistency with the previous and next orbits is often 0.05°. A few orbits have larger uncertainties as a result of data gaps and other issues forcing fallback to the onboard attitudes. Slowly changing systematic errors, especially in pitch, are sensitive to solar beta angle and gyro bias drift uncertainties, but these effects are expected to be less than a few hundredths of a degree.

c. Sensor model and instrument alignment

Geolocation is also affected by the accuracy with which the instrument field-of-view directions are modeled, and each instrument is expected to have some offset in mounting relative to the spacecraft axes. The errors of each instrument in this respect must be analyzed individually.

For TMI, a conical scanning radiometer, a detailed analysis of the instrument alignment offsets was made, as reported in Kroodsma et al. (2017). An offset of the TMI rotation axis relative to the spacecraft Z axis was estimated as −0.08° in pitch and −0.08° in roll. With a nominal 130° active scan, the starting azimuth angle relative to the +X axis was estimated to be −64.36° for the high-frequency feedhorn FOV, and −63.91° for the 10-GHz feedhorn FOV. The nominal FOV offset from the spin axis was 49°; however, better values were found as 49.28° for the high-frequency feedhorn. The 10-GHz feedhorn further showed somewhat different values for the vertical and horizontal polarizations, of about 49.4° and 49.5°, respectively. One set of latitude and longitude coordinates for the 10-GHz channel is provided in the version 8 format, and uses a mean half-cone angle of 49.45°. The geolocation shift for the different V-pol–H-pol cone offsets is about 1 km relative to the mean, which is small relative to the 10-GHz beamwidth (>40 km). However, separate incidence (satellite local zenith) angles are reported for these two channels, since this has some effect on expected brightness temperatures for detailed models. In addition, version 8 incorporates an estimated time delay of about 0.2 s between the 10-GHz and multichannel feedhorn data sampling times, during which time the spacecraft moves over a kilometer along the ground track. These new sensor model parameters are used for version 8 reprocessing.

For the PR and VIRS instruments, new alignment offsets were not considered necessary for reestimation. Previously used prelaunch alignments continue to be used. This decision may add some additional uncertainty (well within requirements) in the geolocation for those sensors.

d. Computations

Several computational improvements for the version 8 reprocessing are noted here. The Geolocation Toolkit software, which was prepared for the GPM mission, is being used in place of the code originally developed for the TRMM science data processing. Many of the basic computations are done for each pixel location rather than including approximations for spacecraft motion and rotation during each scan period.

An error was found in the TRMM code for the sun position calculation, which affected the sun direction by up to about 5° by the end of the mission. Thus, all the ancillary data involving the sun direction are updated significantly for the version 8 reprocessing.

4. TMI emissive antenna correction

a. Counts to brightness temperature

The TMI instrument was a total power radiometer that used two external blackbody targets to achieve radiometric calibration, as illustrated schematically in Fig. 3. The cold radiometric calibration point was the cold-sky mirror (CSM), which captured the cosmic microwave background (CMB) brightness temperature (Tcold) of space, and the hot radiometric calibration point was a microwave absorber blackbody target “hot load.” These calibration targets were viewed during each revolution period (~2 s) of the rotating feeds, to yield a linear inverse transfer equation (i.e., antenna temperature Ta as a function of the digitized radiometer output, counts).

Fig. 3.
Fig. 3.

Block diagram of the two-point calibration radiometer.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

Because the measured Ta was the convolution of the main reflector antenna pattern with the brightness temperature over a spherical surface, it was necessary to apply an antenna pattern correction (APC) to yield the desired Earth scene brightness. Thus, the APC was applied to the Ta to remove the antenna effects of cross polarization and spillover using
e1
where was the antenna temperature of the ith channel after removing the effects of the spillover and the cross polarization; Ta was the corresponding measured copolarized antenna temperature; was the corresponding measured cross-polarized antenna temperature; and Ci, Di, and Ei were the corresponding channel-dependent antenna pattern correction coefficients (measured prelaunch).

b. Emissive antenna

During prelaunch radiometric calibration, a comprehensive thermal vacuum test was performed that simulated the on-orbit environment, but during this test, only the antenna feed was used (i.e., the parabolic reflector was not part of the calibration process). Thus, the first time that the radiometer “viewed the reflector” was on orbit, and during TMI turn on, it was discovered (Wentz et al. 2001) that the TMI main reflector was slightly emissive.

An imperfect (slight emissivity) reflector partially reflects the Earth scene (Tscene) and also emits an unwanted temperature bias that is added to the scene temperature as
e2
where Tscene is the antenna temperature (without emissive reflector), (1 − εi) is the parabolic mirror power reflectivity, εi is the reflector emissivity, Tphy is the reflector physical temperature, and the product (εi) × Tphy is the emission brightness temperature bias. The implementation approach, to remove this emissive reflector effect, has been applied after the APC (correcting for the spillover and the cross polarization). Note that the order of applying the APC and the emissive reflector has a negligible effect on brightness temperature.

Thus, the channel emissivity values were derived by Wentz et al. (2001) during the postlaunch calibration/validation period by comparing measured and theoretical Tb values using a theoretical Radiative Transfer Model (RTM), which was subsequently updated (Wentz 2015). However, for version 8, the Central Florida Remote Sensing Laboratory (CFRSL) independently derived the channel emissivity values based on its analysis of a deep-space maneuver (DSM) that was performed in 2015 (Alquaied and Jones 2017). This more rigorous approach is solely based on radiometric measurements of the known cosmic microwave background, and it avoided uncertainties associated with the use of an RTM and/or intercomparison with another radiometer.

c. Approach

1) Deriving the emissivity coefficients

In 2015, TRMM performed four DSMs to provide “end to end” radiometric calibration by first causing a yaw = 90° and then entering into inertial hold mode, which caused the spacecraft to roll by 360° during one orbit period. As a result, the scanning parabolic reflector antenna viewed space, which is a homogeneous CMB brightness temperature of 2.7–3.2 K over the applicable frequency range. For purposes of calibration, there were three TMI antenna beams to be considered (as shown in Fig. 4), namely, the conically rotating main beam (MB; black arrow), the corresponding main reflector spillover beam (SB; red arrow), and the fixed cold-sky beam (CSB; green arrow). To characterize the emissive antenna, it was necessary that all three beams simultaneously view the cosmic microwave background.

Fig. 4.
Fig. 4.

Orientation of TMI antenna beams during the deep-space calibration roll maneuver. The black arrow is the main beam, the red one the reflector spill-over beam, and the green one the cold sky mirror beam.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

During the DSM, TMI antenna beams went through four different phases as seen in Fig. 4. The first (Fig. 4a) illustrates the spacecraft yawed by 90°, thereby causing the scanning MB to view Earth while the SB and the CSB were pointed to space. The second (Fig. 4b) shows the spacecraft rolled by ~75°, which caused the MB and SB to point to space but now the CSB intersected Earth. The third phase (Fig. 4c) shows the spacecraft rolled by ~200°, which caused the MB and CSB to point to space, but now the SB intersected Earth. Finally, the fourth phase occurred (Fig. 4d) when the spacecraft rotation was ~290° and at that moment all the beams were simultaneously pointed to space.

The antenna temperature shown in Fig. 5 follows the four phases of the DSM. In phase 1, TMI is similar the normal operation mode, where the Ta of channel 1 (10.65 V) ranges between land (280 K) and ocean (180 K). In the second phase, where the CSB pointed to hot Earth, the calibration procedure was corrupted and negative Ta resulted. In the third phase, the CSB departed Earth and pointed to space, so the calibration returned to normal, but now the SB pointed to Earth. As the spacecraft rotation continued, the SB eventually scanned off Earth, and the Ta decreased with time, as seen in Fig. 5b. Finally, when all three beams were pointed to space, the Ta reached a minimum value (scans: 2700–2750) as shown in Fig. 5b. As the spacecraft roll continued, the MB eventually returned to Earth, and the brightness temperature increased.

Fig. 5.
Fig. 5.

(a) Antenna temperature during four phases of the DSM, and (b) the expanded phases 3 and 4.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

During this short period (~100 s) of the fourth phase, the minimum measured antenna temperature was the result of all antenna beams viewing the known CMB brightness plus the unknown MB emissive reflector contribution of brightness; thus, the channel emissivity coefficient was expressed as
e3
where 〈〉 is the mean antenna temperature of channel i after removing the spillover portion, Tcold,i is the CMB temperature for channel i, Tphy is the main reflector physical temperature, and ηi is the prelaunch antenna spillover coefficient as given in Ji et al. (2008).

Because TMI did not have a temperature sensor on the main reflector (MR), it was necessary to use an external source to estimate the Tphy over the orbital cycle that was required in (3). For this purpose, another conical scanning satellite radiometer, the GPM Microwave Imager, was used (Draper et al. 2015). GMI had a reflector antenna of similar construction that was instrumented to measure the reflector physical temperature. Even though these reflectors were provided by different vendors and flew on different spacecraft, there their thermal properties were more similar than different. For example, both reflectors were basically graphite shells with vacuum-deposited aluminum (VDA) coating on the front face and multilayer insulation blankets on the rear surface. Also, the radiometer mounting location on the top decks of the respective spacecraft were very similar, and both antennas were spinning at similar rates. Because both the TRMM and GPM satellites flew in low-inclination, non-sun-synchronous orbits, this resulted in a reflector thermal cycle of equilibrium hot temperature during daylight, followed by an exponential decay to cold equilibrium during eclipse. Also, the dynamic (day–night) reflector temperature orbital cycle was modulated by mean temperature changes of ~50 K peak to peak, as the solar beta angles changed with seasons. Thus, because of the many similarities, it was reasonable to expect similar orbital Tphy cycles by matching the orbital solar beta angles.

Moreover, the derivation of emissivity coefficients relied more on knowledge of the dynamic temperature pattern over the orbit cycle (delta temperature) rather than the mean Tphy value. For example, if the mean Tphy estimate were in error by ±15 K, then the resulting emissivity value would change only by ±5%. Because the emissivity values were ~3%, this Tphy uncertainty would result in a delta emissivity of ±0.0015, which was a negligible impact on the emissive reflector Ta correction.

2) Physical temperature during normal operation

After determining the emissivity values, the Tphy for each orbit was estimated using the 10.65-V channel ocean brightness temperatures. This channel was selected because the ocean brightness was best known for this channel using the intercalibrated RTM (Biswas et al. 2010). The reflector physical temperature was empirically derived to minimize the single difference between the measured and modeled brightness temperatures for this channel using the procedure given in Fig. 6 using
e4
Note that the environmental input parameters for the RTM were from the National Oceanic and Atmospheric Administration’s (NOAA’s) Global Data Assimilation System (GDAS), gridded in 1° latitude/longitude boxes. Finally, the physical temperatures were sorted and saved based on the solar beta angle and the phase from orbit midnight (related to the orbit eclipse time).
Fig. 6.
Fig. 6.

Flowchart of the procedure used to estimate TMI MR physical temperature during normal operation.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

3) MR physical temperature table

The main reflector physical temperature is a common parameter used by all channels to obtain the emissive reflector Tb correction. An example of the derived Tphy orbital cycle for a solar beta angle = −43.5° is displayed as time since entering eclipse (x axis), as shown in Fig. 7. At time equal to zero, the satellite enters into the nighttime of the orbit; thus, the solar heating is blocked and the reflector cools rapidly. After ~30 min, the temperature reaches a minimum value, and as the satellite enters daylight the temperature rises exponentially. This is a typical example of the once (orbit)−1 thermal cycle, which is modulated by seasonal solar heating (beta angle) that raises and lowers the maximum, minimum, and average temperatures. Further, there are frequent, periodic, short-term thermal features, caused by shadows from the spacecraft structure and solar array, that block the solar heating during 5–10-min periods during the daylight portion of the orbit. While this pattern is difficult to model, the derived Tphy orbital cycle is very repeatable (within a few kelvins) given the beta angle and the phase of the orbit midnight (related to the orbit time since eclipse).

Fig. 7.
Fig. 7.

The physical temperature of the TMI’s MR for a solar beta angle = 43.5°.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

Therefore, the derived Tphy has been characterized as a 2D matrix, for use in the TMI counts-to-Tb algorithm, and results (shown in Fig. 8) are the average of 11 yr of data (2003–13). For this matrix the rows are beta angle and the columns are the phase of the orbit midnight. The dynamic range of Tphy is ~50 K over a 46-day period determined by the orbit precession rate.

Fig. 8.
Fig. 8.

The physical temperature of the TMI MR.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

4) Correction Summary

Thus, improved estimates of TMI emissivity coefficients are used for the TMI emissive reflector antenna temperature correction in the legacy processing of the TMI version 8 brightness temperature product. After characterizing the reflector emissivity, an algorithm was applied to calculate the main reflector physical temperature based upon the 10.65-V channel single difference of observed and theoretical brightness temperatures.

The 1B11 version 8 Tb values are quite similar to the version 7 results with differences being generally <±1 K in the mean. Nevertheless, the version 8 dataset is improved significantly because of the rigorous analysis methods employed for the emissive reflector correction that provides a more defensible, physically based correction to the science community users and a significant improvement in the Tb calibration stability, which is crucial for the climate records that are generated from TMI.

Finally, there is one additional correction that has been incorporated into version 8 that was not described. This deals with solar intrusion into the hot load target and the mitigation algorithm that has been developed (Alquaied et al. 2017). It was discovered that there were periodic solar intrusions into the hot load, which caused a small radiometer calibration error. The solution provided a correction of the hot load physical temperature by a lookup table. This and other minor changes will be documented in the future TMI algorithm theoretical basis document (ATBD).

5. TMI level 1 calibration changes

As discussed in section 4a, the TMI calibration occurs on a scan-by-scan basis, and in version 7 and previous versions, the TMI antenna temperature (Ta,i) of each Earth view pixel (i) is calibrated using the linear equations
e5
e6
e7
where A is the inverse transfer function gain, B is the offset; Ci is the Earth view count; Ch and Cc are scan-averaged hot load and cold load counts, respectively; Tcold is the CMB brightness; and Th is scan-averaged hot load physical temperature.

On orbit, the physical temperature of the receiver follows a quasi-sinusoidal variation over the orbital period that also causes the receiver gain to vary. This is not an issue because the frequent calibration removes this systematic gain change over time; however, because of the radiometer noise-equivalent delta Tb (NEDT), the independent scan-by-scan calibration introduces a random scan to scan gain error, which creates alternating intensities of Ta. Further, because the NEDT improves by 1/(number of calibration samples)1/2, this effect varies channel by channel. For example, the 10.6-GHz TMI channel has only 8 calibration samples for each scan, while the 85-GHz channel has 16 samples.

Thus, the channel NEDT of each scan results in a nonstable calibration along the satellite track. Consider the top panel of Fig. 9, which exhibits random transient gain fluctuations along the gain curve of an 18-GHz H channel for a typical TMI orbit. The true receiver gain is actually changing smoothly over time, but the NEDT is introducing these random gain fluctuations, which may produce random scan-by-scan Earth scene Ta fluctuations as much as ±0.5 K along the track. However, the false gain jumps between scans 700 and 800 and between scans 1150 and 1400 are due to the solar intrusions into the hot load and RFI warm intrusions into the cold load. The anomalous gain jumps are discussed later in this section.

Fig. 9.
Fig. 9.

TMI calibration gain of TMI 18-GHz H channel. (top) Version 7 single, and (bottom) version 8 multiscan calibration with hot load and cold load corrections.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

On the other hand, for version 8 TMI a multiscan calibration procedure is applied. The algorithm first examines ±200 scans of a particular scan to determine whether the calibration state of the scan is normal. If it is normal, the algorithm uses nine scan-averaged (current scan plus four scans backward and forward) Th, Ch, and Cc to compute the gain and offset. The bottom panel of Fig. 9 shows that the noise is significantly reduced from the gain curve due to the multiscan calibration. Abnormal scans are flagged and corrected using appropriate methods. For this particular orbit, scans 700–800 have hot load contaminations as a result of solar intrusion, and scans 1150–1400 have warm intrusions onto the cold load, which are corrected in the version 8 TMI calibration algorithm.

Further, Ta differences between multiscan calibration and single-scan differences are shown in Fig. 10. Typically, scan noise ranges between −0.5 and 0.5 K and can be averaged out for a large array of scans. The larger differences in this figure are due to the correction of warm intrusions onto the cold-sky mirror, which have a larger impact over colder ocean scenes.

Fig. 10.
Fig. 10.

TMI 18-GHz H channel Ta differences between versions 8 and 7.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

Transmitters near the TMI bands, especially at the 10.65- and 19.35-GHz TMI frequencies, can provide RFI that results in artificially higher TMI Earth view measurements. The version 8 TMI 1B algorithm flags RFI-contaminated Earth pixels using a method based on spectral differences between the channel of interest and a combination of neighboring channels:
e8
where i represents the channel index to the channel of interest and j is the channel index to all other channels with a different center frequency. The equation given above can be simplified as
e9
Table 1 provides the table of coefficients in (9).
Table 1.

TMI generalized RFI index coefficients.

Table 1.

The ΔTb(i) generated based on (9) are compared to a set of thresholds based on a land mask file. If it is greater than the thresholds, then the Earth sample is flagged (value = 2) as an RFI-contaminated pixel with some exceptions, such as stormy sea, snow, and convective cell. If Tb > 320 K, then the RFI flag is set to 3. The thresholds are presented in Table 2.

Table 2.

Recommended thresholds for flagging RFI in a real-time algorithm.

Table 2.

Figures 11 and 12 demonstrate the RFI flagging of TMI swath 1 and TMI swath 2 in the version 5A TMI Base file of orbit 84890. The red (number = 3) indicates Tb > 320 K. Figure 13 shows the 10-GHz Tb for the same area.

Fig. 11.
Fig. 11.

RFI flag for 10-GHz Earth samples.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

Fig. 12.
Fig. 12.

RFI flag for 19-GHz Earth samples for the same area shown in Fig. 11.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

Fig. 13.
Fig. 13.

The Tb of 10 GHz for the same orbit as in Figs. 11 and 12.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

The typical warm intrusions onto the cold load include RFI and moonlight. These can increase the cold load count significantly. Such intrusions also cause a much larger-than-normal gradient across samples of a scan and across scans along the track. The version 8 algorithm compares along-scan and cross-scan gradients to the normal maximums and determines the calibration state of the scan by combining the two results. If a scan is flagged as a warm intrusion, then the algorithm looks forward and backward to find the nearest valid scans to do a linear interpolation. The algorithm uses previous and postorbits to perform the calibration near orbit boundaries. Figure 14 demonstrates the effectiveness of the TRMM version 8 processing by comparing the scan mean cold load counts with corrections (bottom panel) and without corrections (top panel).

Fig. 14.
Fig. 14.

TMI cold load count of TMI 18-GHz H channel. (top) Version 7 single-scan calibration, and (bottom) version 8 multiscan calibration with RFI correction.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0166.1

6. Conclusions

Considerable effort has been expended to improve the accuracy of the TMI brightness temperatures. This effort is important, as TMI acts as the calibrator for microwave radiometers that flew only during the TRMM era and for creating long-term climate data records. Such high-quality and consistent brightness temperatures also ensure consistent precipitation retrievals.

The final TRMM project reprocessing of the data ensures consistency with the GPM data suite. Further, it incorporates this eighth TRMM reprocessing into the GPM V05 data suite; as such, TRMM data will be continually reprocessed as algorithms improve in GPM. Most importantly, the incorporation of TRMM into GPM ensures that GPM has a consistent data suite of 20-plus yr.

Last, consistent retrievals from the beginning of TRMM onward allow for the creation of consistent merged radiometer products for the entire period. Upon completion of the eighth reprocessing of TRMM, the TRMM Multisatellite Precipitation Analysis (TMPA) product (3 hourly at 0.25° × 0.25° resolution) will be replaced with the Integrated Multisatellite Retrievals for GPM (IMERG) product (half hourly at 0.1° × 0.1° resolution) for the entire TRMM data period.

Inclusion in the GPM data suite is an indication of NASA’s commitment to ensuring the best and most consistent global precipitation retrievals back to the beginning of the TRMM era.

Acknowledgments

The authors wish to acknowledge the work of Dr. David Draper of Ball Aerospace for his TMI RFI analysis. This work was the basis for the new version 8 RFI flagging approach. The work done by the University of Central Florida coauthors was funded under NASA Grant NNX16AE35G. Considerable analysis was done by all members of the GPM Intercalibration Working Group in the area of TMI algorithm improvement. The authors wish to thank Dr. John Kwiatkowski for providing an overview of level 1 TRMM PR changes, and Ms. Jeanne Beatty for her extensive formatting and editing of this paper.

REFERENCES

  • Alquaied, F., and W. L. Jones, 2017: An emissive antenna correction for the Tropical Rainfall Measuring Mission Microwave Imager. 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 2104–2107, https://doi.org/10.1109/IGARSS.2017.8127398.

    • Crossref
    • Export Citation
  • Alquaied, F., R. Chen, and W. L. Jones, 2017: Hot load temperature correction for the Tropical Rainfall Measuring Mission Microwave Imager (TMI). 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 2108–2111, https://doi.org/10.1109/IGARSS.2017.8127399.

    • Crossref
    • Export Citation
  • Berg, W., and Coauthors, 2016: Intercalibration of the GPM radiometer constellation. J. Atmos. Oceanic Technol., 33, 26392654, https://doi.org/10.1175/JTECH-D-16-0100.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bilanow, S., and S. Slojkowski, 2006: TRMM on-orbit performance reassessed after control change. Extended Abstracts, 19th Int. Symp. on Space Flight Dynamics, Kanazawa City, Japan, JAXA, ISTS 2006-d-35, 3 pp., https://pps.gsfc.nasa.gov/Documents/Bilanow_TRMM_Pointing_Control_ISTS_2006-d-35_a.pdf.

  • Biswas, S. K., K. Gopalan, W. L. Jones, and S. Bilanow, 2010: Correction of time-varying radiometric errors in TRMM Microwave Imager calibrated brightness temperature products. IEEE Geosci. Remote Sens. Lett., 7, 851855, https://doi.org/10.1109/LGRS.2010.2050135.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Draper, D. W., D. A. Newell, D. S. McKague, and J. R. Piepmeier, 2015: Assessing calibration stability using the Global Precipitation Measurement (GPM) Microwave Imager (GMI) noise diodes. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 8, 42394247, https://doi.org/10.1109/JSTARS.2015.2406661.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ji, Y., S. Bilanow, and E. Stocker, 2008: NASA TRMM TMI level 1B algorithm, version 1. TRMM Science and Data Information System Algorithm Theoretical Basis Doc., 40 pp., https://pps.gsfc.nasa.gov/Documents/L1B11ATBD_V1.pdf.

  • Kroodsma, R., S. Bilanow, Y. Ji, and D. McKague, 2017: TRMM Microwave Imager (TMI) updates for final data version release. 2017 IEEE International Geoscience and Remote Sensing Symposium Proceedings, IEEE, 1216–1219.

  • Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson, 1998: The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Oceanic Technol., 15, 809817, https://doi.org/10.1175/1520-0426(1998)015<0809:TTRMMT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kummerow, C., J. Simpson, O. Thiele, W. Barnes, A. T. C. Chang, E. Stocker, and R. F. Adler, 2000: The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit. J. Appl. Meteor., 39, 19651982, https://doi.org/10.1175/1520-0450(2001)040<1965:TSOTTR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Skofronick-Jackson, G., and Coauthors, 2017: The Global Precipitation Measurement (GPM) mission for science and society. Bull. Amer. Meteor. Soc., 98, 16791695, https://doi.org/10.1175/BAMS-D-15-00306.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stocker, E., 2016: File naming convention for precipitation products for the Global Precipitation Measurement (GPM) mission. NASA GSFC PPS_610.2_P550, Version 1.4.6, 22 pp., https://pps.gsfc.nasa.gov/Documents/FileNamingConventionForPrecipitationProductsForGPMMission.pdf.

  • Takahashi, N., and T. Iguchi, 2004: Estimation and correction of beam mismatch of the precipitation radar after an orbit boost of the Tropical Rainfall Measuring Mission satellite. IEEE Trans. Geosci. Remote Sens., 42, 23622369, https://doi.org/10.1109/TGRS.2004.837334.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 2015: A 17-year climate record of environmental parameters derived from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager. J. Climate, 28, 68826902, https://doi.org/10.1175/JCLI-D-15-0155.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., and D. Draper, 2016: On-orbit absolute calibration of the Global Precipitation Measurement Microwave Imager. J. Atmos. Oceanic Technol., 33, 13931412, https://doi.org/10.1175/JTECH-D-15-0212.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., P. Ashcroft, and C. Gentemann, 2001: Post-launch calibration of the TRMM Microwave Imager. IEEE Trans. Geosci. Remote Sens., 39, 415422, https://doi.org/10.1109/36.905249.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save