Algorithm Overview:

    The 1B21 calculates the received power at the PR receiver input point from the Level-0 count value which is linearly proportional to the logarithm of the PR receiver output power. To convert the count value to the input power, extensive internal calibrations are applied, which are mainly based upon the system model, temperature dependence of model parameters and many temperature sensors attached at various locations of the PR. Periodically the input-output characteristics are measured using an internal calibration loop for the IF unit and later receiver stages. To make an absolute calibration, an Active Radar Calibrator (ARC) is placed at Kansai Branch of CRL and overall system gain of the PR is being measured every 2 months. Using the transfer function based on the above internal and external calibrations, the PR received power is obtained. Note that the value assumes that the signal follows the Rayleigh fading, so if the fading characteristics of a scatter is different, a small bias error may occur (within 1 or 2 dB).
The other ancillary data in 1B21 include:

    -    Locations of Earth surface and surface clutter (range bin number). Those are useful to identify if the echo is rain or surface.
    -    System noise level: One value per angle bin. This is the reference noise floor which is used to extract echo power from the "total" received
        power in 1B21 (echo + noise).
    -    Over-sample data: In order to improve the accuracy of surface echo measurement, and to obtain a better vertical rain profile, every 125-m
        data are available at a near-nadir rain region (up to 7.5 km) and around surface (+/-10 deg. scan angles).
    -    Minimum echo flag: a measure of the existence of rain within a beam. There are several confidence levels and users may select up to what
         confidence level they treat as rain.
    -    Bin storm height: The maximum height at which an echo exists for a specific angle bin.
    -    Land/ocean flag and Topographic height

    The 1C21 calculates the effective radar reflectivity factor at 13.8 GHz without any propagation loss (due to rain or any other atmospheric gas) correction (Zm). Therefore, the Zm value can be calculated just by applying a radar equation for volume scatter with PR system parameters. The noise-equivalent Zm is about 21 dBZ. Through the subtraction of the system noise, the Zm value as small as 16 or 18 dBZ are still usable although the data quality is marginal. In 1C21, all echoes stored in 1B21 are converted to "dBZ" unit. This is not relevant for "non-rain" echo; however, this policy is adopted so that the 1B21 and 1C21 product format should be as close as possible except for the following points:

    -     Radar quantity is Zm in dBZ unit instead of received power (dBm).
    -     Data at echo-free range bins judged in 1B21 are replaced with a dummy value.

File Format

    The file content description for 1B21/1C21 can be obtained from TRMM Homepage. It is available at: http://www.eorc.nasda.go.jp/TRMM/product/pl/l1_frmt_e.pdf
ECS Metadata Elements - http://www.eorc.nasda.go.jp/TRMM/product/pl/ECS_me.pdf
PS Metadata Elements - http://www.eorc.nasda.go.jp/TRMM/product/pl/PS_me.pdf

Comments on PR Level 1 products

1.    Calibration accuracy

    The TRMM Precipitation Radar (PR) has been working without any problem since the first turn-on of the PR power in the beginning of December 1997. The initial checkout of the PR was completed by NASDA and CRL at the end of January 1998. The overall calibration of the PR including the transmit and receiving antenna pattern measurements were made by using an ARC. It was concluded that the ARC calibration results are reasonable and consistent with a calculation from PR system parameters. Also the ocean surface sigma-0 obtained by the PR has been found to be quite consistent with those observed from previous airborne and satelliteborne scatterometers.

2.     Sensitivity

    The minimum detectable Zm (corresponding to the noise-equivalent received power) improved from 23.3 dBZ (based upon the specifications requirement) to 20.8 dBZ as determined from the pre-launch ground test and from the orbit test . This is mainly due to the increased transmit power and the decrease of the receiver noise figure. Actually the rain echo power is measured from the subtraction of the system noise power from the total
receiver power (rain echo power + system noise power). The accuracy of rain echo power can be characterized by the effective signal-to-noise ratio (S/N), that is the ratio of mean to standard deviation of rain echo power. By considering these facts, the actual minimum detectable Zm can be considered to be about 16-18 dBZ after the detailed statistical calculation. The effective signal-to-noise ratio (S/N) of 3 dB is obtained when Zm is 17 dBZ.

3.     Discrimination of rain from surface clutter

    It is generally very difficult to discriminate rain echo from surface clutter especially in mountainous regions. An algorithm has been implemented which analyzes the radar echo range profile very carefully to determine the boundary between rain and surface echoes. The result has been reflected into the surface location related variables described in Item 6. Even though, there is a very small possibility that a surface echo is treated as a rain echo
(and in mountainous regions, clutter position when it rains can happen to become too high in a very rare occasion). Please be careful when you use the PR Level-2 data to study rain structure in mountainous regions. Strong echoes near the surface are likely surface clutter and should be excluded from rain analysis.

4.     Surface clutter from the coupling between nadir-direction antenna sidelobe and strong surface radar cross-section (NRCS)

    It has been found that the echo strength from nadir direction is sometimes extremely strong, which exceeds the anticipated value in the PR design. This seems to occur wet and flat land areas rather than ocean. Even dry dessert regions, the NRCS seems very strong in some cases. In such cases antenna sidelobes directed to nadir receive surface echoes. When main beam is off-nadir, the timing of such nadir-surface clutter can contaminate the
rain echo. In "PR STATUS2", a warning flag is set ON when nadir surface echo (an the nadir angle bin, #25) exceeds a pre- determined threshold. When it is ON, please be careful about the echo at the range bin number the same as the Bin_surface_peak at nadir (angel bin number 25).

5.     Discrimination of rain echo from noise

    In order to help users utilization of the data, the 1B21 product contains the "Minimum Echo Flag" which indicates the existence of rain in the clutter free range or in the clutter range. Since thermal noise, rain echo and resulting thermal noise plus rain echo follow Rayleigh fading, the PR received echo is a result of the averaging 64 number of independent samples. The averaged value still have small fluctuations of about 0.7 dB to 1 dB, depending on
signal-to-noise ratio. In order not to miss weak echo which is sometimes useful to study rain structure, etc, the threshold to set the flag = rain possible is currently about 90% value of the cumulative distribution of thermal noise. This means quite a large fraction of data having "rain possible" flag is only thermal noise. Since this rain/no-rain discrimination is sometimes affected by the surface clutter at especially mountainous area. In the clutter region, rain/no-rain discrimination often misidentifies clutter as rain. Minimum Echo Flag includes clutter flag. There are five levels in the Minimum Echo Flag; 0, 10, 20, 11, and 12:

    0    =  no rain (Echoes are very weak),
    10   =  rain possible but maybe noise (Some weak echoes above noise exist in clutter free ranges),
    20   =  rain certain (Some strong echoes above noise exist in clutter free ranges),
    11   =  rain possible but maybe noise or surface clutter (Some weak echoes exist in possibly cluttered ranges), and rain possible but maybe clutter
    12   =  (Some strong echoes exist in possibly cluttered ranges). Therefore please be careful in using the Minimum Echo Flag except 0 and 20.

6.     Information concerning the surface location in 1B21 and 1C21.

    The following variables are newly added in 1B21 and 1C21 products.

    a.     Range bin number of Ellipsoid (binEllipsoid)
    b.     Range bin number of clutter free bottom (binClutterFreeBottom)
    c.     Range bin number of mean DID (binDIDHmean)
    d.     Range bin number of top of DID (binDIDHtop)
    e.     Range bin number of bottom of DID (binDIDHbottom)

    As you can imagine from the name of each variable, those represent range bin numbers corresponding to the height from the Earth ellipsoid, which may be useful to analyze a range profile of PR received power or radar reflectivity factor.

7.     Bin_surface_peak and oversample data

    In PR 1B21, the data called Bin_surface_peak indicates the range bin number at which PR received power has the maximum within a range window centered at the range bin number determined from a Digital Elevation Model (DID). In most cases, the Bin_Surface_Peak gives the correct location corresponding to the location of actual surface. There may be small number of cases where Bin_surface_peak is wrong. One possibility is cases where
DID is in error, and the other is cases where rain echo is so strong so that surface echo is masked by the rain echo. We expect those cases are rare, but please keep in mind those may occur with a small probability. The over-sample data are recorded onboard based on the location of surface echo peak
detected by an onboard surface tracking function. Since this tracker may be locked-off in mountainous regions, there are cases where oversample data are recorded outside the location of surface echo. In such cases the oversample data may not be useful because it may not be used for improving the accuracy of surface echo power or for detailed study of vertical storm structure. The difference between the location of surface echo estimated by
the onboard tracker (Note 1) and Bin_surface_peak is a measure of the goodness of oversample data in terms of its covering region in the radar range profile. Note 1: The surface echo location estimated onboard (Y) can be obtained from "Bin_start_oversample" data. Let X be Bin_start_oversample,
Y = X + 60 or + 61 (angle bins between 20 and 30) and Y = X + 4 or +5 (between 11 and 19 and between 31 to 39). We can not judge either 60 or 61 (or 4 or 5) from 1B21 itself, however.

8.     Interference from other radio services around 13-14 GHz

    There have been several cases where PR suffered from interference from other radio services, mainly from satellite tracking and control stations using 13-14 GHz bands. The probability is very small, and the impact to TRMM mission appears to be negligible. In a typical interference case, the noise level increases a few to several decibels over entire range bins for a very short period (one or two scans). In such a case, PR sensitivity to detect weak
echo is degraded accordingly.

Planned Improvements

    Routine monitoring of PR performance and periodical ARC calibration are being conducted.

    Depending on the drift of PR system parameters, Calibration factors may be updated in future.

References

T. Kozu, T. Kawanishi, K. Oshimura, M. Satake, H. Kumagai; TRMM precipitation radar: calibration and data collection strategies, Proc. IGARSS'94, 2215-2217, Pasadena, 1994.

M. Satake, K. Oshimura, Y. Ishido, S. Kawase, T. Kozu: TRMM PR data processing and calibration to be performed by NASDA, Proc. IGARSS'95, 426-428, Florence, 1995.