1

Detection and Determination of Channel Frequency Shift in AMSU-A Observations

Cheng-Zhi Zou and Wenhui Wang

IGARSS 2011, Vancouver, Canada, July 24-28, 2011IGARSS 2011, Vancouver, Canada, July 24-28, 2011

NOAA/NESDIS/Center for Satellite Applications and Research

(Thanks Y. Han and Y. Chen at JCSDA for their CRTM calculation support)

2

Background

Weighting functions for AMSU-A. All weighting functions are corresponding to nadir or near-nadir observations.

AMSU-A: 1998-present on NOAA-15 through NOAA-19 and MetOp-A, NASA Aqua

AMSU-A observations are being assimilated into NWP models for accurate weather prediction in most weather centers in the world

AMSU-A observations are being assimilated into climate reanalysis systems to constrain model climate

AMSU-A observations are merged with MSU by different research groups to generate atmospheric temperature time series for climate change monitoring

In all these applications, channel frequency values are specified to be

the pre-launch measurements

Bias corrections of unknown error sources are conducted before AMSU-A data are being assimilated into NWP and reanalysis models

This study identify one of these error sources using inter-satellite bias analysis method

AMSU-A Orbit Information

Satellites Launch Date LECT at lunch

NOAA-16

SEPT 2000

1400 Ascending

NOAA-15

MAY 1998

0730 Descending

NOAA-17

JUNE 2002

1000 Descending

NOAA-18

MAY 2005

1400 Ascending

MetOp-A October 2006

0930 Descending

Local Equator Crossing Time of the Descending Orbits of the NOAA and MetOp-A satellites

SNO Datasets

For polar orbiting satellites, SNO events are generally found over the polar region

Use Cao’s (2004) orbital method to find SNO events

4

Schematic viewing SNO and its locations

Examples of SNO Inter-Satellite Biases

5

Channel 6 of MetOp-A minus NOAA-18 Channel 6 of NOAA-15 minus NOAA-18

6

k j

Radiance Error Model for SNO Matchup K and J

kkkkLk ZRRR 0,

jjjjLj ZRRR 0,

SNO Radiance Error Model

jjkkL ZZRRR 0

kj jk RRR 000

j

Remove relative mean inter-satellite biases

Remove non-uniformity in inter-satellite biases

Remove instrument temperature signals

7

Effect of Calibration Non-linearity

Channel 6 of MetOp-A minus NOAA-18 Channel 6 of MetOp-A minus NOAA-18

Before Inter-Calibration After SNO Inter-Calibration

Lapse Rate Climatology

Average over the 700S The averaged lapse rate around 350 hPa being steeper in winters (July) than in summers (January).

Time series with winter values being at the negative side of the summer values when the frequency shift is positive (weighting function peaking higher than prelaunch measured), and the other way around for negative frequency shift.

NOAA-15 should have a positive frequency shift

Channel 6 Measurement

NOAA-15 Minus NOAA-18

9

Pre-launch Measured Frequencies for AMSU-A Channel 6

Measured Channel Frequency (Specification =54400 for all satellites)

NOAA-15 54399.53

NOAA-16 54399.78

NOAA-18 54400.97

MetOp-A 54400.07

Frequency characteristics for AMSU-A Channel 6 from Mo [1996; 2006; 2007]. Units are in MHz.

Measured frequency differences between different satellites are within 0.5 MHz.

These errors are so small that they wouldn't result in noticeable Tb differences between satellites (0.01K)

Practically, these measured channel frequencies can be considered as the same for different satellites

The shift is a post-launch error Differences for all pairs: 0.5 MHz

Methods to Determine the Actual Channel Frequency Use NOAA Joint Center for Satellite Data Assimilation (JCSDA) Community Radiative Transfer Model (CRTM) to simulate NOAA-15 observations at its SNO sites relative to NOAA-18

Use NASA MERRA reanalysis surface data and atmospheric profiles (temperature, humidity, ozone, cloud liquid water, trace gases etc.) as inputs to the CRTM

MERRA data were interpolated into the N15-N18 SNO sites before being used by CRTM

Select different frequency shift values (df) in the simulation experiments

Analyze Tb(N15, df) = Tb(N15, fm + df) - Tb(N15, fm)

fm : Measured Channel Frequency Valuedf: Frequency Shift

Experimental Results

Comparisons between simulations and observed N15-N18 SNO data confirms a positive frequency shift in the NOAA-15 channel 6 relative to its measured frequency value Observed SNO time series over

the Antarctic between NOAA-15 and NOAA-18

Simulated Tb (N15, df)

12

Determine the Final Channel Frequency Value

 

 

Examine Tb’, which is the Tb differences between NOAA-15 and NOAA-18 at their SNO sites when NOAA-15 Tb is adjusted by its simulated frequency shift

We expect the seasonal cycles in Tb’ disappear when df equals to the actual channel frequency shift’

The seasonal cycles can be measured by the amplitude, which should be equal to zero for df=actual channel frequency shift

)18(),15()15(' NTdfNTNTT bsbbb

dfo = 36.25±1.25MHz

fa = fm+ dfo = 54435.73±1.25 MHz

13

Impact on SNO Time Series

Channel 6 of NOAA-15 vs NOAA-18Before Frequency adjustment

Channel 6 of NOAA-15 vs NOAA-18After NOAA-15 Frequency adjustment

14

Method is developed to detect and determine the post-launch channel frequency shift in AMSU-A observations onboard polar orbiting satellites

NOAA-15 channel-6 frequency shift is determined

Methods are expected to be applicable to other satellites and other channels, but analysis has to be done for each channel, since all channels have different lapse rate climatology

Call for impact experiments on NWP accuracy improvement at JCSDA; if positive, we need to work on more channels

Also call for provisional parameters for future AMSU-type instruments, allowing calculating the frequency shift after launch

Conclusion