SEARCH

SEARCH BY CITATION

Keywords:

  • satellite data assimilation;
  • calibration;
  • validation;
  • MWTS;
  • MWHS;
  • IRAS;
  • MWRI

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

FY-3A, launched in May 2008, is the first in a series of seven polar-orbiting meteorological satellites due to be launched by China's Meteorological Administration in the period leading up to 2020. The FY-3A payload includes four instruments of particular interest for numerical weather prediction (NWP): microwave temperature and humidity sounders, a microwave imager, and an infrared sounder. The main features of these instruments are described. Data from the calibration–validation phase of the FY-3A mission were introduced into the ECMWF Integrated Forecasting System in order to assess the data quality and the influence of the data on analyses and forecasts. An analysis of first-guess departures has shown the data to be of good quality overall. Several issues with instrument performance and ground segment processing have been identified. The most serious of these are: uncertainties in the temperature sounder passbands on-orbit, orbital biases in the infrared instrument affecting the highest peaking channels, and scan biases in the microwave humidity sounder. Variational bias correction partially corrects for these errors, but more work remains to be done to correct the problems before the full benefit of the data is realised. In observing system experiments, the FY-3A instruments, both individually and as a package, show considerable skill when added to observation depleted control experiments. When added to a full observing system, the impacts are neutral to slightly positive, as expected. These initial results are encouraging and build confidence that the following series of FY-3 instruments will be widely used in NWP data assimilation systems. Copyright © 2011 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

China has successfully launched and operated both geostationary and polar-orbiting satellites (Li, 2001; Meng, 2004) which have been named FengYun, meaning wind cloud, often shortened to FY-N. An odd value of N denotes the polar orbiting series, whereas an even N denotes the geostationary series. Following the launch of China's first polar-orbiting satellite (FY-1A) in 1988, China has launched a series of four further polar orbiters (FY-1B/C/D and FY-3A) and five geostationary satellites (FY-2A/B/C/D/E). Both programmes will continue over the next decade and an ambitious schedule of launches is currently planned, accommodating increasingly sophisticated sensors for operational meteorology.

FY-3A is the preparatory platform for a subsequent series of six polar-orbiting satellites (FY-3B to FY-3G) currently planned for launch between 2010 and 2020. FY-3A was launched from the Taiyuan Launching Centre on 27 May 2008. The payload of FY-3A comprises a suite of eleven instruments (Dong et al., 2009; Zhang et al., 2009). Of particular interest for NWP data assimilation (DA) are the three instruments which make up the Vertical Atmospheric Sounder System (VASS): the Microwave Temperature Sounder (MWTS), the Microwave Humidity Sounder (MWHS), and the Infrared Atmospheric Sounder (IRAS). These cross-track scanning instruments are similar, but not identical, in specification to the Microwave Sounding Unit (MSU), the Advanced Microwave Sounding Unit-B (AMSU-B, recently replaced by the US/European Microwave Humidity Sounder, or MHS) and the High Resolution Infrared Sounder (HIRS) carried originally on the US Polar Orbiting Environmental Satellite System (POES) (Goodrum et al., 2000). Also of interest for NWP is the 10-channel Microwave Radiation Imager (MWRI), a conical scanning instrument similar in specification to the Advanced Microwave Scanning Radiometer (AMSR) (Kawanishi et al., 2003).

Since it is a preparatory platform, it is important that the performance of the FY-3A sensors is assessed carefully, in order that any deficiencies in the design or on-orbit operation of the instruments can be addressed in future instruments. As part of a comprehensive calibration/validation programme, radiances measured by the suite of VASS sounders as well as the MWRI have been compared to radiances modelled from ECMWF short-range forecast fields, using radiative transfer modelling. This approach is now common practice as the high accuracy of NWP fields, coupled with accurate radiative transfer calculations, permits the detection of systematic errors in the satellite-measured radiances (Bell et al., 2008). For example, for temperature sounding radiances, modelled brightness temperatures can reveal systematic errors of a few tenths of a Kelvin in measured brightness temperatures. For moisture-sounding channels the sensitivity is lower, but errors of around 1 K can be detected using this technique.

This study describes an initial evaluation of FY-3A data using the European Centre for Medium-range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS). This evaluation has focused on assessing data quality through an inspection of first-guess departure fields (differences in observed and simulated radiances based on short-range forecasts) and associated statistics. The data have also been assessed through Observing System Experiments (OSEs), in which data from the four FY-3A instruments have been introduced into baseline (observation depleted) and full-system experiments, in order to assess the impact on NWP analyses and forecasts.

In section 2, a more complete description of the FY-3A instrument characteristics is given. In section 3, the assessment of data quality is presented based on an analysis of first-guess departures. The observing system experiments are described in section 4 and some conclusions are drawn in section 5.

2. The FY-3A instruments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

2.1. FY-3A

FY-3A is a sun-synchronous polar-orbiting environmental satellite with an orbital inclination of 98.8° and equatorial crossing time of 1005 LST. The design life of FY-3A is three years. The FY-3A spacecraft carries eleven instruments, but in this study we focus on only the four instruments of FY-3A of most interest for NWP DA, i.e. the MWTS, the MWHS, the IRAS and the MWRI.

2.2. Microwave Temperature Sounder (MWTS)

Until the recent advent of advanced IR sounding instruments, microwave temperature sounding data from high-performance radiometers was the single most important satellite data type in NWP DA systems (English et al., 2004). Microwave temperature sounding data, by providing accurate information for the analysis of mass fields, is still a key component of NWP DA systems (Cardinali, 2009).

The FY-3A MWTS is a four-channel cross-track scanning radiometer similar in specification to the MSU carried on the US POES series of satellites (NOAA-8 to NOAA-14). The channel characteristics are shown in Table I. The main reflector of the instrument is rotated through a single measurement cycle once every 16 s to give, for each scan line, 15 scene fields of view together with a view of cold space and the on-board warm calibration target, used to perform a two-point radiometric calibration. The resulting swath width is 2250 km.

Table I. FY-3A MWTS, and equivalent AMSU-A, channel characteristics and assumed observation errors (R).
ChannelFrequencyFrequencyFrequency§BandwidthMWTSAMSU-ARR
number (GHz)(GHz)(GHz)(MHz) NEΔTNEΔTMWTS(AMSU-A)
     (K)(K)(K)(K)
  • *

    equivalent AMSU-A channel number in parentheses.

  • specification.

  • pre-launch measurement.

  • §

    optimised.

1 (3)*50.3050.26250.262180 0.500.2110.030.0
2 (5)53.596±0.11553.60153.6562×170 0.150.130.430.35
3 (7)54.9454.98155.020400 0.140.180.520.35
4 (9)57.2957.34057.373330 0.190.160.550.35

The reflector aperture is 10.7 cm in diameter which gives rise to a footprint diameter of 62 km at nadir. From Table I, it can be seen that radiometric sensitivity, expressed as a noise equivalent brightness temperature (NEΔT) is approximately 0.14–0.19 K for the channels 2–4, similar to the performance from AMSU-A equivalent channels. These values have been derived from on-orbit data. Based on previous experience (English et al., 2004) with introducing new sounding instruments (both infrared and microwave) into operational systems, it is likely that the largest impacts on forecast accuracy from the VASS suite will result from the use of the MWTS radiances. It is also to be expected from the outset that the positive impact of the MWTS instrument will be less than that obtained from AMSU-A, with a more complete set of temperature-sounding channels spanning the troposphere and lower stratosphere.

Table I also shows the observation errors assumed in the OSEs described later in section 4. These errors determine the weight given to the MWTS observations, relative to the background (or first-guess) information, in these experiments. These values are typically larger that the true errors associated with the measurements, arising from random radiometric noise as well as residual uncorrected biases. In practice, these errors are determined empirically by choosing a value which optimises forecast performance. Possible reasons why this inflation is necessary include the need to compensate for observation-error correlation and the over-specification of background errors (relative to the true background errors), but this is still an active research topic. The observation errors for MWTS, and the other FY-3A instruments, have been derived by scaling the optimised error estimates for equivalent ATOVS channels by the ratio of the standard deviations of first-guess departures (i.e. FY-3A/ATOVS). The process of optimising these estimates independently for MWTS, through multiple OSEs, is computationally expensive and beyond the scope of the initial evaluation reported here.

The specifications of MWTS instruments for subsequent FY-3 missions will be progressively improved. For example FY-3C, which is currently planned for launch in 2012, will carry a 13-channel MWTS instrument similar in specification to AMSU-A.

2.3. Microwave Humidity Sounder (MWHS)

Microwave sounding channels spanning the water vapour absorption line at 183 GHz have been shown to provide important information for the analysis of mid- to upper-tropospheric humidity in global DA systems (Andersson et al., 2007). To date, these data have been provided by sensors on the NOAA POES platforms (AMSU-B and MHS) and more recently the European MetOp-A platform (MHS).

The MWHS on board FY-3A is a 5-channel instrument similar in specification to AMSU-B/MHS, with channels in the frequency range 150–183 GHz. The channel characteristics are shown in Table II. MWHS has an aperture of 14 cm, giving a nominal field of view of 15 km at nadir. MWHS differs from AMSU-B in that it includes a dual-polarisation channel at 150 GHz (v, h) (channels 1 and 2 respectively) rather than including a channel at 89 GHz. MWHS measures 98 fields of view in each 2700 km cross-track scan in contrast to AMSU-B/MHS which measures 90 scenes over a 2250 km scan.

Table II. FY-3A MWHS channel characteristics and assumed observation errors (R).
ChannelFrequencyBandwidthMWHSMHSMHSRR
number(GHz) (pol)(MHz)NEΔTChannelNEΔTMWHSMHS
   (K)number(K)(K)(K)
1150 (v)10000.920.84
2150 (h)10000.9
3183.31±15001.130.602.262.0
4183.31±310000.940.702.462.0
5183.31±720000.951.062.392.0

It is planned that the MWHS will be further improved for FY-3D through the inclusion of three channels in the 118 GHz O2 band. These channels will provide information on temperature, water vapour and cloud fields which complements that available from traditional sounding (50–60 GHz) and imaging channels (19, 22, 37 and 89 GHz).

2.4. Infrared Atmospheric Sounder (IRAS)

Infrared sounders have been used for NWP DA since the launch of the first HIRS instrument on NIMBUS 6 in 1975. Originally designed to be the primary temperature sounding instrument on the TOVS/ATOVS platforms, the practical difficulties of reliably screening the IR measurements for cloud radiative effects, as well as the reduced coverage due to cloud contamination, has reduced the impact of filter-based IR radiometric data. Consequently microwave sounders have had a more significant positive impact in NWP DA systems. The introduction of advanced high resolution infrared instruments based on high performance interferometers (IASI; Challon et al., 2001) or grating spectrometers (e.g. AIRS; LeMarshall et al., 2006) offering high spectral resolution, wide spectral coverage, excellent radiometric performance and hence enhanced information in the vertical has elevated the importance of IR sounding data for NWP DA applications. (Hilton et al., 2009; Collard and McNally, 2009). The treatment of cloud-affected radiances, both in the infrared and microwave, remains a priority for research at NWP centres and further advances in forecast impact are anticipated.

As a first step towards an FY-3 advanced IR sounding capability, the InfRared Atmospheric Sounder (IRAS) (a HIRS/3-like instrument) is the primary IR sounder for FY-3A. IRAS has a total of 26 channels, the first 20 of which are similar to HIRS/3 channels while the six additional channels enable IRAS to measure aerosols, carbon dioxide columns, and cirrus clouds. The channel specifications of the IRAS instrument are given in Table III. The instrument is a cross-track scanning radiometer with a ground footprint 17 km in diameter at nadir. Calibration is achieved once every 40 scan lines by a two-point calibration based on deep space views and warm target views.

Table III. FY-3A IRAS channel characteristics and assumed observation errors (R).
ChannelFrequencyBandwidthNEΔNHIRSNEΔNRR
number(cm−1)(cm−1)(IRAS)Channel(HIRS)(IRAS)(HIRS)
   (mW m−2sr−1cm)number(mW m−2sr−1cm)(K)(K)
166934.0013.00
2680100.8020.67
3690120.6030.50
4703160.3540.311.100.60
5716160.3250.210.990.60
6733160.3660.240.800.60
7749160.3070.200.830.75
8802300.2080.15
9900350.1590.10
101030250.20100.15
111345500.23110.200.900.91
121365400.30120.201.141.22
131533550.30130.0060.800.50
142188230.01140.0031.200.60
152210230.01150.004
162235230.01160.004
172245230.01170.002
182388250.01180.002
192515350.01190.001
2026601000.002

2.5. Microwave Radiation Imager (MWRI)

The MWRI is a conical-scanning microwave imager operating at five frequencies in the range 10–89 GHz, each with dual polarisation, giving 10 channels in total. The channel specifications are shown in Table IV. The MWRI has an aperture of 90 cm, giving a frequency-dependent ground footprint in the range 12–80 km over a swath width of 1400 km. The rotation of the main reflector of the FY-3A MWRI has led to imbalances in the FY-3A platform and this has resulted in an observing strategy in which the MWRI is only spun up and operated for short and intermittent periods. The assessment of the data in the following section has been based on some of the available data segments, obtained during the periods 10–17 September 2008 and 10–20 October 2008.

Table IV. FY-3A MWRI channel characteristics.
ChannelFrequencyBandwidthNEΔTAMSR-EBandwidthNEΔT
number(GHz)(MHz)(MWRI)Channel(MHz)(AMSR-E )
(v, h) pairs  (K)number (K)
1,210.651800.53,41000.6
3,418.72000.55,62000.6
5,623.84000.87,84000.6
7,836.59000.59,1010000.6
9,1089.046001.011,1230001.1

3. Data quality assessment

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

3.1. First-guess departure statistics

An overview of the data quality for the key channels of the four FY-3A instruments, in terms of the standard deviations of the first-guess departures after variational bias correction (VarBC) and quality control, is given in Figure 1. Also shown for comparison are the equivalent statistics for the corresponding MetOp-A instruments and, for the case of the MWRI instrument, the equivalent AMSR-E statistics. The statistics have been obtained over one month. These statistics measure the fit of the ECMWF model to the observed radiances and give a good early indication of data quality, as any gross errors in the data would be manifested as a large spread in the first-guess departures. The VarBC scheme uses the same predictors for the FY-3A instruments as for the analogous ATOVS instruments (and other microwave imagers in the case of MWRI). The sounding channels of IRAS, MWTS and MWHS use eight predictors: a constant global offset, four thickness predictors (1000–300 hPa, 200–50 hPa, 50–5 hPa and 10–1 hPa) and first, second and third powers of the satellite view angle to compensate for cross-track biases. For MWRI, the predictors used are: a global offset ocean surface skin temperature and wind speed, total column water vapour, and first, second and third powers of the satellite view angle. After a spin-up period of approximately 10 days, the bias correction coefficients were relatively stable over periods of 1–3 months. The quality control for the FY-3A instruments has been based on the equivalent ATOVS instruments where possible (e.g. in the case of IRAS and MWHS). Where there is a significant difference in channel sets (e.g. MWTS compared to AMSU-A), existing quality control schemes have been adapted.

thumbnail image

Figure 1. Standard deviations of bias-corrected first-guess departures for FY-3A and MetOp-A/AQUA equivalent instruments and channels. Statistics are derived from used data for the period from 28 August to 28 September 2008: (a) MWTS and AMSU-A, (b) MWHS and MHS, (c) IRAS and HIRS, and (d) MWRI and AMSR-E.

Download figure to PowerPoint

Figure 1(a) shows a comparison for MWTS/AMSU-A. The standard deviations for MWTS channels 3 and 4 (0.25–0.26 K) are significantly larger than those for AMSU-A channels 7 and 9 (0.2–0.21 K). There are several possible reasons for this: firstly, as discussed later in section 3.2, residual localised biases caused by uncertainties in the channel passband centre frequencies would be expected to increase the global standard deviation. Secondly, there appears to be evidence of nonlinearity in the MWTS instrument which would also be expected to increase the variance in the first-guess departures.

Figure 1(b) shows the comparison for MWHS/MHS for the 183 GHz water vapour sounding channels. The standard deviations for the MWHS channels are around 15% larger than the MHS channels, most probably due to the complex scan biases which are not dealt with well in the current VarBC scheme (section 3.4).

The comparison for IRAS/HIRS is shown in Figure 1(c). For the temperature sounding channels (4–7, 703–749 cm−1), the standard deviations of the first-guess departures for IRAS are 50–70% larger than the corresponding HIRS channels, as a result of the orbital biases discussed in section 3.3.

The comparison of the MWRI with AMSR-E, valid for the period 10–17 September, is shown in Figure 1(d). MWRI compares well with AMSR-E for most channels, and is similar in quality to AMSR-E with standard deviation of first-guess departures up to 25% larger than the equivalent AMSR-E channels. For the channels at 23.8 GHz and 36.5 GHz, the MWRI first-guess departure statistics are slightly better than AMSR-E.

3.2. MWTS passband shift

The passband characteristics for the MWTS instrument, based on both initial instrument specifications and pre-launch measurements are shown in Table I. Pre-launch measurements showed significant differences in the passband centre frequencies for some channels relative to the design-specified passband centre frequencies, e.g. for channel 4 the difference is 50 MHz. The accurate specification of passbands is required for generating the coefficients for the fast radiative transfer model (RTTOV9; Saunders, 2010). In the absence of measurements of the passband shape, as is the case here for MWTS, idealised rectangular passbands are assumed which are centred on the chosen passband centre frequency. Line-by-line calculations (Liebe, 1989; Liebe et al., 1992, 1993) using this passband are then used to derive coefficients for the fast-regression-based model (Eyre, 1991).

An initial inspection of brightness temperature fields and histograms for MWTS observations (Figure 2(a, b)) showed a significant offset relative to those for the MetOp-A AMSU-A equivalent measurements. For example, for MWTS channel 4, the peak in the distribution, at ∼209 K (associated with measurements in the Tropics), is shifted by +2 K relative to the AMSU-A observations. Although these shifts could result from a number of possible causes, e.g. errors in the antenna temperature to brightness temperature conversion; radiometric calibration errors or radiometer nonlinearities), a closer inspection shows that the latitudinal variation of the shift (relative to AMSU-A) indicates it is largely due to a shift in the passband of MWTS channel 4. Similar effects are also observed for channels 2 and 3. The strong surface and cloud effects on the channel 1 radiances make the signal more difficult to detect. This passband shift is manifested by an apparent airmass dependence of the first-guess departure field (Figure 2(c, d)). The absence of this airmass dependence in the analogous AMSU-A channel (channel 9; Figure 2(f)) indicates that the error is unlikely to be caused by model bias or bias in the radiative transfer model. Estimates of the passband centre frequency were subsequently refined and were shown to reduce the standard deviation of the first-guess departures (Figure 2(e)) before VarBC. These studies also highlighted a small degree of nonlinearity in the MWTS radiometer. The refined estimates of the passband centre frequencies are shown in Table I.

thumbnail image

Figure 2. (a) and (b) show the observed brightness temperatures for MWTS (channel 4) and MetOp-A AMSU-A (channel 9) respectively, for the 12 h 0000 UTC analysis cycle on 17 September 2008. (c) shows the first-guess departure field for MWTS channel 4 for the specified passband centre frequency (57.29 GHz), and (d) shows the same based on the pre-launch measured passband centre frequency (57.339 GHz). (e) shows the first-guess departures based on an optimised estimate of the passband centre frequency (57.373 GHz), and (f) shows the equivalent plot for AMSU-A channel 9. All first-guess departures are before variational bias correction has been applied. Assumed passband centre frequencies (ν0) and standard deviations of the first-guess departure fields are also indicated.

Download figure to PowerPoint

3.3. IRAS biases

Figure 3 shows first-guess departures for IRAS channels 1 and 4 (temperature sounding channels at 669 and 703 cm−1 respectively), both before and after bias correction, alongside bias-corrected MetOp-A HIRS first-guess departures for equivalent channels. The spread in first-guess departures is larger for the IRAS data than the equivalent HIRS data. This is particularly evident for channel 1. Several features are evident from these plots: firstly, anomalous first-guess departures are obtained for a particular section of one orbit from the IRAS instrument (between 60°W and 120°W); secondly, there are areas of positive bias in the southern polar regions in the IRAS observations which are not evident in the HIRS data; and finally there appears to be biases related to the albedo of the underlying surface, indicative of spectral leakage for the filter associated with the highest peaking channel (1). These biases are more severe for the highest peaking IRAS channels (channels 1–3) and are locally as large as 10 K (for channel 1). These biases appear to be related to the orbitally varying illumination of the satellite by the sun, and are correlated with solar-induced heating of the satellite.

thumbnail image

Figure 3. First-guess departures for FY3-A IRAS before (left) and after (centre) bias correction, and MetOp-A HIRS (right). First-guess departures are shown for the temperature sounding channels that are currently used in operations (channels 4 to 7, from top), at cycle 2008091700.

Download figure to PowerPoint

3.4. MWHS scan biases

Figure 4 shows the scan biases for the MWHS, compared to those for equivalent MetOp-A MHS channels. The dominant component of the scan bias, for channels 1-4, is a complex modulation across the swath with a peak–peak amplitude in the range 0.5 to 2 K. The low-order polynomial (in scan angle) bias predictors currently used in the ECMWF VarBC scheme are unable to deal adequately with this form of cross-track scan bias and consequently biases are evident even after bias correction. Such biases have been observed in other instruments, for example in NOAA-19 AMSU-A channel 7 and present difficulties for the existing ECMWF bias correction scheme. Possible solutions to this problem would be to implement a scan-position-dependent offset in the pre-processing of the radiances, using values derived from this study, or to implement a scan-position-dependent predictor in VarBC.

thumbnail image

Figure 4. First-guess departure statistics and bias correction versus scan position for FY-3A MWHS and equivalent MetOp-A MHS channels. Statistics are based on 9 days of used data (20–28 September 2008). Red indicates the mean departure after bias correction, with the green lines indicating the standard deviation. Blue is the mean bias correction.

Download figure to PowerPoint

4. Observing System Experiments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

4.1. Introduction

OSEs are commonly used as a means of assessing the value of new observational datasets in NWP DA systems. Recent examples include the work at ECMWF to assess the influence of the main humidity observing systems (Andersson et al., 2007) and the impact of microwave imager data in NWP (Kelly et al., 2008; Geer et al., 2008). An approach commonly used to assess the value of new data sources is to add data to an observation-depleted observing system. In such a baseline system, analysis and forecast errors are larger than might be expected from a full observing system. Consequently, the improvements in analysis and forecast accuracy are larger than would be the case for a full observing system. The larger signals in forecast skill measures thereby facilitate a comparison of a new data type with other similar observation types. This type of experiment provides a check that there are no serious problems with a new observational data type. These baseline experiments, of course, do not give a realistic representation of the forecast improvement to be expected when the new data are added to a full system. Full system experiments are conducted to assess the impact of new data types when analysis errors are small, and are normally carried out in the later stages of pre-operational testing. In the context of the evaluation of FY-3A data, the purpose of full system experiments is to check that the new data do not cause any measurable degradation to forecast performance. Both baseline and full system experiments are described in sections 4.2 and 4.3 below.

4.2. Baseline experiments

Initial baseline experiments were run at low resolution (T159, equivalent to approximately 125 km grid resolution) to assess the impact of the individual FY-3A instruments relative to that from the equivalent NOAA/MetOp instruments against an observation-depleted control experiment. All experiments described here used the operational 4D-Var DA scheme. Radiance observation operators were based on the RTTOV-9 fast radiative transfer model. Subsequently, higher-resolution (T511) baseline experiments were run to assess the impact of the combined VASS suite of instruments (MWTS, IRAS, MWHS) relative to the ATOVS suite of instruments on MetOp-A. Both sets of experiments are described below.

4.2.1. Baseline experiments at T159 resolution: Assessing the impact of individual instruments

The control chosen for these experiments included a depleted observing system in which only conventional observations, satellite observations from the F13 and F15 Special Sensor Microwave/Imager (SSM/I) instruments as well as data from six global positioning system radio occultation (GPSRO) sensors were included in the analysis. SSMI data were assimilated as radiances using the All-Sky assimilation scheme as described by Bauer et al. (2010). In this scheme observations are assimilated in clear, cloudy and precipitating areas. The conventional observations included temperature, pressure, wind and humidity measurements from the global network of surface stations, radiosondes, dropsondes, aircraft, buoys and ships. Feature track winds are also included in this baseline configuration.

GPSRO bending angle data are assumed to have very low measurement uncertainties in the altitude range 10–35 km as the measurements are derived from phase delay (time) measurements which are traceable to highly accurate primary measurement standards. The systematic component of the measurement uncertainties are therefore very small and the observations are assimilated without bias correction. This effectively anchors the VarBC system, ensuring that long-term drifts in model bias are prevented. Recent OSEs have shown that GPSRO satellite provide a powerful constraint on the large-scale analyses (S. Healy, personal communication), and results in forecast impacts close to that from IASI in the Southern Hemisphere (SH). Overall, in the SH GPSRO data provides around 50% of the impact of a full observing system relative to a conventional observations only control experiment. The control configuration chosen here is therefore a reasonable choice. MWTS, MWHS, IRAS and equivalent MetOp-A ATOVS instruments were added separately to this control experiment:

  • Control: conventional observations + SSMI + GPSRO

  • Baseline 1: Control + FY-3A MWTS

  • Baseline 2: Control + MetOp-A AMSU-A (5,7,9)

  • Baseline 3: Control + FY-3A MWHS

  • Baseline 4: Control + MetOp-A MHS

  • Baseline 5: Control + FY-3A IRAS

  • Baseline 6: Control + MetOp-A HIRS

The experiments were run for the period 20 August to 1 October 2008 and the resulting forecast verified against operational analyses. Observation errors for the FY-3A instruments were determined by scaling the observation errors for the equivalent ATOVS instruments by the ratio of the standard deviations of the first-guess departures from MWTS and AMSU-A. Only channels 5, 7 and 9 from MetOp-A AMSU-A were assimilated to provide as near to a like-for-like comparison as possible.

As a summary of the impacts, the resulting anomaly correlation scores (for 500 hPa geopotential height) and RMS errors for humidity for these experiments are summarised in Figure 5. Relative to the control, the FY-3A MWTS shows significant forecast skill impact, amounting to 70% of the impact of AMSU-A at day 4 in the SH. The impact for IRAS is positive, but smaller than for MWTS, amounting to 20% of the impact of MetOp-A HIRS at day 4. This may be partially attributable to the residual biases in the bias-corrected IRAS data. The MWHS shows very similar impact to that from MetOp-A MHS, with measurable improvements to humidity forecast to day 3. The first-guess departure statistics of Baseline3 from 28 August to 28 September are shown in Figure 6 to demonstrate the impact of assimilating FY-3A MWHS on the first-guess departure fits for radiosonde observations in the Tropics and SH and dropsonde fits in the Northern Hemisphere (NH). First-guess departure fits for humidity are generally improved for the 200–700 hPa levels in the Tropics and SH. Fits are unchanged in the NH.

thumbnail image

Figure 5. Anomaly correlation for 500 hPa geopotential height in (a, b) the Northern Hemisphere and (c, d) the Southern Hemisphere for an observing system depleted control experiment (red) as well as experiments in which either FY-3A (green) or MetOp-A (blue) sounding data have been added. (e, f) show root mean square errors for 500 hPa relative humidity for NH and SH.

Download figure to PowerPoint

thumbnail image

Figure 6. First-guess and analysis departure statistics for relative humidity (%) from dropsondes in the NH (top) and radiosondes in the Tropics (middle) and SH (bottom), for 28 August to 28 September 2008. Baseline + FY3A MWHS is in black and the Baseline experiment is in red.

Download figure to PowerPoint

4.2.2. Baseline experiments at T511 resolution: Assessing the impact of the VASS instrument suite

The control experiment here used data from a single AMSU-A in order to provide some constraint on the background fields. Data from NOAA-18 were used as this satellite was in an orbit complimentary to the FY-3A platform. This meant that FY-3A data were influencing the analysis at different locations to the NOAA-18 data, with the NOAA-18 data providing some control on the quality of the analysis. The following experiments were run over the period 20 July to 1 November 2008:

  • Control: conventional observations + NOAA 18/AMSU-A + ozone data + scatterometer data

  • Baseline A: Control + MetOp-A AMSU-A (channels 5, 7, 9) + MHS + HIRS

  • Baseline B: Control + FY-3A (MWTS + MWHS + IRAS)

The results, in terms of the anomaly correlation of the forecasts in the NH and SH over the period 30 July to 1 November 2008 are shown in Figure 7. In the NH the impact of the FY-3A data is very similar to that from the MetOp-A instruments. In the SH the impact is, for forecast ranges to day 7, positive for the FY-3A data and a significant fraction of that obtained from MetOp-A.

thumbnail image

Figure 7. Anomaly correlation for 500 hPa geopotential height in (a) the Northern Hemisphere and (b) the Southern Hemisphere for an observing system depleted control experiment (red) and experiments in which either FY-3A (green) or MetOp-A (blue) sounding data have been added.

Download figure to PowerPoint

First-guess departure statistics for radiosonde observations in the NH, Tropics and SH for the Baseline B experiment from 20 July to 30 October 2008 are shown in Figure 8 to demonstrate the impact of assimilating FY-3A VASS on the short range forecasts of humidity. FY-3A data significantly improves the first-guess humidity fields in the Tropics and SH. Impacts are smaller in the NH.

thumbnail image

Figure 8. First-guess and analysis departure statistics for radiosonde humidity in the NH (top), Tropics (middle) and SH (bottom) for 20 July to 30 October 2008. Baseline experiment + FY3A VASS is in black and the baseline experiment is in red.

Download figure to PowerPoint

4.3. Full system experiments

A full system experiment was run over the period 20 July to 1 November 2008, using a T511 configuration of the operational model. The suite of FY-3A sounding instruments (MWTS, MWHS and IRAS) were added to the full ECMWF system. The OSEs carried out were therefore:

  • Full System 1: CY35R2 full system experiment

  • Full System 2: CY35R2 full system experiment + FY-3A (MWTS + MWHS + IRAS)

The forecast verification results are summarised in Figure 9. Normalised differences in RMS errors in geopotential are selected as these indicate the impact of the new observations on the large-scale mass fields. The component of the FY-3A suite most likely to have an effect on these verification measures is the MWTS. Two of the three actively assimilated channels of the MWTS (MWTS-2 and -3) have weighting functions peaking at or below 200 hPa and hence the choice of 200 hPa geopotential as a verification measure permits an assessment of the value of MWTS. In the SH, forecast impacts are neutral for all forecast ranges for both 200 hPa and 500 hPa geopotential errors. In the NH, the picture is more mixed with small (0.5–1.0%) degradations at forecast ranges to T+24 hours followed by small improvements, albeit at the limits of statistical significance, in the longer range. This improvement seems coherent from T+3 to T+7 days and in both 200 hPa and 500 hPa verification measures. These near-neutral results are encouraging and represent a first step in the exploitation of FY-3A data in NWP systems. The impact of the data is likely to improve as biases in the data are eliminated, quality control improved and error tuning is optimised.

thumbnail image

Figure 9. Normalised differences in the root mean squared forecast error between the (Full + FY3A) experiment and the Full system experiment (Full system minus (Full system + FY-3A); i.e. positive values indicate a positive benefit from FY-3A data) for the 0000 UTC forecast of the 500 hPa geopotential height for (a) the NH and (b) the SH. Verification is against operations, and the period is 10 August–1 November 2008 (93 cases); the experiments were run from 20 July to 1 November 2008. Error bars indicate 90% significance intervals from a t-test. (c, d) are as (a, b), but for 200 hPa geopotential height.

Download figure to PowerPoint

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Data from the FY-3A VASS instruments as well as the microwave imager (MWRI), obtained during the July–November 2008 calibration–validation phase of this preparatory mission, have been introduced into the ECMWF Integrated Forecasting System. This has permitted an assessment of data quality through an analysis of first-guess departure statistics, a comparison with equivalent ATOVS instruments, as well as through observing system experiments. The use of NWP model fields for the calibration and validation of satellite missions is becoming standard practice as the global coverage and high accuracy provided by NWP models have been proven to be a powerful tool in the diagnosis of systematic biases in the data, and this has again been the case with the assessment presented here. The assessment of the data through OSEs provides, in addition, valuable information on the impact of the data on NWP analysis and forecast quality.

In the OSEs presented here, the individual VASS instruments (MWTS, MWHS and IRAS) were able to show significant positive impact, equivalent to a significant fraction of the impact from the equivalent MetOp-A ATOVS instruments when assimilated in a like-for-like configuration. As a package, the VASS instruments show impact equivalent to a significant fraction (e.g. ∼30% at day 5 in SH 500 hPa geopotential height anomaly correlation) of that obtained from the MetOp-A ATOVS suite, even when added on top of a control experiment with one AMSU-A. When added to a full OSE, the impact of the FY-3A data is neutral to slightly positive, a result which is encouraging and indicates that there are no major problems with the data from FY-3A from an NWP perspective. Overall the results from these OSEs, using data from the calibration–validation phase of this preparatory mission, are a very promising start to the FY-3 programme.

Perhaps more importantly, this assessment has highlighted several issues where further work needs to be done to refine the instrument design and pre-launch testing as well as the ground processing systems, in order to further improve the quality of FY-3 data for FY-3A and subsequent missions. Specifically, accurate measurements of the passband filter shape and centre frequency stability should be made over the full range of on-orbit operating conditions. Similarly the pre-launch radiometric measurements should include a characterisation of nonlinearities and radiometric noise performance over the full range of on-orbit conditions. Simulations and measurements of field-of-view interference from other instruments or the spacecraft platform would assist in the development of bias correction schemes. Improvements in these areas are perhaps even more critical for the use of the data for climate and reanalysis studies.

One important lesson from this study, also drawn from experience of POES, MetOp and DMSP missions, is that early assessment of the data using NWP systems is a valuable component of the calibration and validation of satellite missions. Early data access ensures that instrument issues can be flagged early and mitigation measures put in place in time for subsequent launches. Useful feedback can also be generated by NWP centres on the impact of changes to the ground processing systems or instrument configuration during the early orbit commissioning phase of a mission. A related benefit for NWP centres is that processing systems are developed and tested at the earliest possible opportunity, hence maximising the benefit extracted from the data.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References

This work was supported by the bilateral cooperation agreement between ECMWF and the China Meteorological Administration (CMA). The work was also supported by the National Natural Science Foundation of China (grant No. 40705037). The Research and Development Special Fund for Public Welfare Industry (Meteorology GYHY(QX)2007-6-9) provided international travel funds. The production of this manuscript was supported by the EUMETSAT NWP SAF programme.

The authors would like to thank a large group of people at ECMWF, and beyond, for help with this work: Blazej Krzeminski for help on the technical aspects of the IFS; Alan Geer for guidance on the all-sky microwave imager assimilation scheme; Marco Matricardi, Sabatino Di Michele, Roger Saunders, Peter Rayer, Pascal Brunel and Amy Doherty for help with the radiative transfer aspects of the work; Dingmin Li for valuable discussions and suggestions during the implementation of FY-3A in IFS; Tony McNally for assistance with data quality checks and observation statistics; Milan Dragosavac for assistance with BUFR encoding; Anne Fouilloux for help with ODB aspects of the work; Mohamed Dahoui and Lars Isaksen for help with data monitoring software; Jan Haseler for valuable system insights; Anabel Bowen and Rob Hine for their patience in providing graphics; and Dick Dee for guidance on variational bias correction.

The authors would also like to acknowledge, with thanks, the contribution from the colleagues who worked on the FY-3A engineering programme in the National Satellite Meteorological Centre of CMA, especially the instrument teams, for their helpful discussions. Finally the authors would like to thank all our colleagues who coordinated the ECMWF–CMA cooperation agreement.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The FY-3A instruments
  5. 3. Data quality assessment
  6. 4. Observing System Experiments
  7. 5. Conclusions
  8. Acknowledgements
  9. References
  • Andersson E, Hólm E, Bauer P, Beljaars A, Kelly GA, McNally AP, Simmons AP, Thépaut J-N. 2007. Analysis and forecast impact of the main humidity observing systems. Q. J. R. Meteorol. Soc. 133: 14731485.
  • Bauer , P, Geer AJ, Lopez P, Salmond D. 2010. ‘Direct 4D-Var assimilation of all-sky radiances. Part I: Implementation’. Tech. Memo. 618. ECMWF: Reading, UK. Available at www.ecmwf.int/publications/library/do/references/list/14.
  • Bell W, English S J, Candy B, Hilton F, Atkinson N, Swadley S, Baker N, Bormann N, Kazumori M. 2008. The assimilation of SSMIS radiances in numerical weather prediction models. IEEE Trans. Geosci. Remote Sensing 46: 884900.
  • Cardinali C. 2009. Monitoring the observation impact on the short-range forecasts. Q. J. R. Meteorol. Soc. 135: 239250.
  • Challon G, Cayla F, Diebel D. 2001. ‘IASI: An advanced sounder for operational meteorology’. In Proceedings of the 52nd Congress of IAF, Toulouse, France.
  • Collard AD, McNally AP. 2009. The assimilation of Infrared Atmospheric Sounding Interferometer radiances at ECMWF. Q. J. R. Meteorol. Soc. 135: 10441058.
  • Dong C, Yang J, Yang Z, Lu N, Shi J, Zhang P, Liu Y, Cai B, Zhang W. 2009. An overview of a new Chinese weather satellite FY-3A. Bull. Amer. Meteorol. Soc. 90: 15311544.
  • English SJ, Saunders R, Candy B, Forsythe M, Collard AP. 2004. ‘Met Office Satellite Data Observing System Experiments’. In 3rd WMO Workshop on the Impact of Various Observing Systems in Numerical Weather Prediction, Alpbach, Austria.
  • Eyre JR. 1991. ‘A fast radiative transfer model for satellite sounding systems’. Tech. Memo. 176. ECMWF: Reading, UK. Available at http://www.ecmwf.int/publications/library/do/references/list/14.
  • Geer AJ, Bauer P, Lopez P. 2008. Lessons learnt from the operational 1D+4D-Var assimilation of rain- and cloud-affected SSM/I observations at ECMWF. Q. J. R. Meteorol. Soc. 134: 15131525.
  • Goodrum G, Kidwell KB, Winston W. 2000. NOAA KLM Users Guide. NESDIS, NCDC: Suitland, Maryland, USA. Available at http://www2.ncdc.noaa.gov/docs/klm.
  • Hilton F, Atkinson NC, English SJ, Eyre JR. 2009. Assimilation of IASI at the Met Office and assessment of its impact through observing system experiments. Q. J. R. Meteorol. Soc. 135: 495505.
  • Kawanishi T, Sezai T, Ito Y, Imaoka K, Takeshima T, Ishido Y, Shibata A, Miura M, Inahata M. Spencer RW. 2003. The Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E); NASDA's Contribution to the EOS for Global Energy and Water Cycle Studies. IEEE Trans. Geosci. Remote Sensing 41: 184194.
  • Kelly G A, Bauer P, Geer A J, Lopez P, Thépaut J-N. 2008. Impact of SSM/I observations related to moisture, clouds, and precipitation on global NWP forecast skill. Mon. Weather Rev. 136: 27132726.
  • LeMarshall J, Jung J, Derber J, Chahine M, Treadon R, Lord SJ, Goldberg M, Wolf W, Liuand HC, Joiner J, Woollen J, Toddling R, Van Delst P, Tahara Y. 2006. Improving global analysis and forecasting with AIRS. Bull. Amer. Meteorol. Soc. 87: 891894.
  • Li Q. 2001. Development of Chinese geostationary meteorological satellite, 2001. Spacecraft Recovery Remote Sensing 22: 1319.
  • Liebe H. 1989. MPM—An atmospheric Millimeter-wave Propagation Model. Int. J. Infrared Millimeter Waves 10: 631650.
  • Liebe HJ, Rosenkranz PW, Hufford G. 1992. Atmospheric 60 GHz Oxygen spectrum: New Laboratory measurements and line parameters. J. Quant. Spectrosc. Radiat. Transfer 48: 629643.
  • Liebe H, Hufford G, Cotton M. 1993. ‘Propagation modeling of moist air and suspended water/ice particles at frequencies below 1000 GHz’. AGARD 52nd Specialists' Meeting of the Electromagnetic Wave Propagation Panel, Palma de Mallorca, Spain.
  • Meng Z. 2004. The Polar Orbit Meteorological Satellite in China. Eng. Sci. 6: 15.
  • Saunders R. 2010. ‘RTTOV Science and Validation Plan’. Met Office: Exeter, UK. Available at http://research.metoffice.gov.uk/research/interproj/nwpsaf/.
  • Zhang P, Yang J, Dong C, Lu N, Yang J, Shi J. 2009. General introduction on payloads, ground segment and data application of Fengyun 3A. Frontiers Earth Sci. China 3: 367373.