NPP ATMS Instrument On-orbit Performance K. Anderson, L. Asai, J. Fuentes, N. George Northrop Grumman Electronic Systems ABSTRACT The first Advanced Technology Microwave Sounder (ATMS) was launched on October 28, 2011 aboard the Suomi National Polar-orbiting Partnership (NPP) spacecraft. ATMS represents the next generation of polar orbiting microwave sounders that are used to create global models of atmospheric temperature and moisture profiles. The ATMS continues the crucial role of legacy AMSU A and B in weather forecasting and long term climate research. This paper presents the on-orbit performance of ATMS, including comparisons of key parameters such as radiometric accuracy and sensitivity (NEDT). Assessments based on data obtained from both operational orbits and during special cal/val maneuvers demonstrate improved data quality relative to the AMSU-A and B. The ATMS design features that contributed to the ATMS performance improvements and benefits (e.g. noise figures, sampling rate, and thermal stability) are also discussed. INTRODUCTION The Advanced Technology Microwave Sounder (ATMS) was developed by Northrop Grumman Electronic Systems under a NASA contract, for the Suomi National Polar-orbiting Partnership (NPP) program, which was launched on October 28, 2011. A nearly identical, functionally equivalent followon unit is presently in system-level integration, for deployment on the JPSS-1. The ATMS was specified to provide essentially the same data products as the prior AMSU-A and MHS instruments (primarily atmospheric temperature and humidity sounding), but with the following enhancements: 1. Faster scan rate and sampling rate, providing Nyquist sampling of temperature sounding (facilitating data fusion with the CrIS) 2. Additional sounding channels 3. Thermal control via spacecraft cold-plate, to provide more flexibility for mounting location 4. Improved torque disturbance compensation 5. Improved reliability and longer life 6. Reduced size, weight and power, relative to the heritage sensor suite The channels and spectral characteristics are shown in Table 1, which also identifies nearest corresponding AMSU-A,B channels. In addition to the above listed improvements, the ATMS provides better NEDT performance, when processed for equivalent spatial resolution. It has also been found that the use of the spacecraft heat rejection system results in improved temperature stability, and consequent radiometric stability. All ATMS functionality and engineering telemetry are nominal and fully consistent with ground test data. The sections below present summary results from assessment of on-orbit data, obtained during the NPP Calibration/Validation phase.
ATMS Chan AMSU-A,B (MHS) Chan Center Freq (GHz) Pre-Detection Bandwidth (MHz) 1 1 23.8 270 QV 2 2 31.4 180 QV 3 3 50.3 180 QH 4 51.76 400 QH 5 4 52.8 400 QH 6 5 53.596±0.115 170 QH 7 6 54.4 400 QH 8 7 54.94 400 QH 9 8 55.5 330 QH 10 9 57.290344 155 QH 11 10 57.290344±0.217 78 QH 12 11 57.290344±0.3222±0.048 36 QH 13 12 57.290344±0.3222±0.022 16 QH 14 13 57.290344±0.3222±0.010 8 QH 15 14 57.290344±0.3222±0.0045 3 QH 16 15,16 (H1) 88.2 2000 QV 17 17 (H2) 165.5 1150 QH 18 20 (H5) 183.31±7.0 2000 QH 19 183.31±4.5 2000 QH 20 19 (H4) 183.31±3.0 1000 QH 21 183.31±1.8 1000 QH 22 18 (H3) 183.31±1.0 500 QH Table 1: ATMS spectral channels and equivalent AMSU-A,B channels. Pol INSTRUMENT STABILITY The critical performance parameters for which stability was assessed were the receiver shelf temperatures, the gain of each channel, and the warm calibration load temperatures. During orbit 164 it was determined that all these parameters were sufficiently stable to meet radiometric performance requirements, but there was still a noticeable warm-up drift. A re-assessment on orbit 182 indicated no noticeable drift, and this data can be used to assess magnitudes of systematic orbital variations. Example results for this orbit are shown in Figure 1. The observed stabilities for all channels were significantly better than had been assumed in the analyses for predicting radiometric calibration accuracy. For example, the channel 3 gain stability is 1. X 10-5 db/sec, versus a predicted 8. X 10-5 db/sec; and the warm load stability is < 0.00035 K/sec, versus a predicted 0.001 K/sec. Consequently, the calibration accuracy assessment below, based on these observed stabilities, shows some improvement over the previous predictions. INSTRUMENT RADIOMETRIC SENSITIVITY The sensitivities of all channels were derived from orbit 182 warm load counts data, by computing the standard deviation of counts relative to a linear regression, and using an average gain value to convert to Noise Equivalent Delta Temperature (NEDT). The values were then scaled up to the value that would correspond to a 300K scene temperature, for comparison to the specified requirements. Representative results are shown in the charts in Figure 2. In these charts, both the ATMS observed NEDT s and the ATMS requirements for channels 1-16 have been divided by three, to allow a direct comparison with AMSU-A. This is because the data sampling interval is a factor of three smaller for ATMS, in both the along-track and cross-track directions. When
ATMS data is averaged to the equivalent AMSU footprint, as is done for fusion with CrIS data, there is thus a factor of three reduction of noise in the final re-sampled SDR. This adjustment for spatial resolution was not done for channels 17-22, since they are compared against the MHS, which has the same sampling and scanning rate as the ATMS. This data demonstrates significant improvements in sensitivity, as would be expected from the more advanced receiver front-end technology. Figure 1: Stability of V-band shelf temperature (channels 3-15), gain for channel 3, and warm calibration load temperature (used for channels 1-15) during orbit # 182.
Figure 2: Sensitivity of ATMS channels, for orbit 182, scaled to 300 K scene, and converted to AMSU equivalent footprints. INSTRUMENT RADIOMETRIC ACCURACY There has not yet been a complete assessment of on-orbit radiometric accuracy, but there is on-orbit data available for updating some of the parameters used in the analytical model. In particular, the gain drifts, the warm calibration load temperatures and drifts, and the observed shelf temperatures. The analytical model includes a direct contribution from gain drifts that can occur over a calibration interval, which had been based on worst-case predicts. Since the on-orbit data indicate extremely stable shelf temperatures, this item in the budget is eliminated. Similarly for warm calibration target temperature drifts. Also, the modeling of residual on-orbit calibration temperature error assumes a nominal warm calibration target temperature. With the on-orbit data available, it is appropriate to substitute the actual measured temperatures into the model. Finally, one can use the quadratic nonlinearity corresponding to actual receiver shelf temperatures, rather than a worst-case over the specified range of temperatures. The results after incorporating these factors in the model are summarized in the chart in Figure 3. This chart indicates the specified ATMS requirements, the predicted worst-case when using a linear calibration algorithm, and predicted worst-case using a quadratic correction term in the algorithm. The three primary sources of calibration error are warm load brightness temperature, cold-space view brightness temperature, and non-linearity of the transfer function. The relative contributions from these three sources depends on the observed scene temperature, and the values reported here are for the worst-case over a range of scene temperatures from 180 K to 300 K. The error estimate for the warm load contribution assumes worst-case modeling of temperature gradients and stray radiation inputs, but they have been recomputed using the actual on-orbit load temperatures. The estimate for cold-space view errors is the same as previously modeled, based on measured antenna sidelobes and worst-case assumptions regarding the uncertainties of the sidelobe contribution.
Figure 3: Predicted radiometric accuracy, using measured on-orbit parameters. The estimated error due to non-linearity is shown for options of either using a linear algorithm for radiometric calibration, which corresponds to the assumptions behind the stated specified requirement, and for the option of using a quadratic correction coefficient, which corresponds to the actual operational SDR algorithm. Previous analyses used the worst-case ground calibration results over the full range of instrument temperatures and redundancy configurations. For the results reported here, we have used the primary redundancy configuration (which is the present configuration), and quadratic coefficients that would correspond to the on-orbit measured shelf temperatures. For the linear algorithm case, the non-linearity contribution is the deviation of the quadratic curve from a straight line; for the quadratic algorithm case, this error contribution is not included. The conclusion is that there are significant margins relative to the requirements, and significant further improvements when incorporating the quadratic term in the calibration algorithm. SCAN-DEPENDENT BIAS One of the phenomena that has long been observed for cross-track scanning radiometers (e.g. AMSU-A) is a radiometric brightness temperature offset that is a systematic function of scan angle. Based on extensive analyses of scene data, attempts have been made to empirically model this error and apply a correction. Although several hypotheses have been considered, the root cause had never been established. During the NPP ATMS calibration/validation phase, pitch-over maneuvers were performed, which allowed a view of cold-space across the entire scan, with virtually no sidelobe intercepts with the earth. This provided valuable data to permit a better characterization of this phenomenon, and for investigating candidate causes. Figure 4 shows examples of this scan bias, obtained from the pitch maneuver of 20 February 2012. Scan angles from -52.725 to +52.725 are in the Earth-View sector, and the 4 samples at 81.675 85.005 are in the Cold Calibration sector. Counts were converted to brightness temperatures using the average gain of each channel, and applying offsets such that: 1. Each QV channel temperature at 0 scan angle is at zero Kelvin 2. Each QH channel temperature at 0 scan angle is equal to twice the temperature at +45
These offsets are not intended to produce absolute actual sky temperatures, but rather to facilitate relative comparisons of the bias effect. Figure 4: Scan-dependent bias, obtained from calibration/validation pitch maneuver, compared to simulated effect of reflector emissivity and exo-atmosphere. It is quite significant that when observing the virtually uniform cosmic background, there is a highly symmetrical variation with scan angle: for the quasi-vertical (QV) channels, it is a good fit to a cosine-square curve, and for the quasi-horizontal (QH) channels it is a good fit to a sine-square curve. See Table 1 for identification of QV and QH channels. The meaning of quasi-vertical is that the polarization is in the scanning plane when viewing nadir, but it rotates during the scan. Similarly, the quasi-horizontal channels are polarized normal to the scanning plane at the nadir scan position, and the polarization rotates during the scan. This is due to the fact that the scanning is implemented by a cross-track scanning reflector that is illuminated by a fixed feed horn. The likely explanation for these observations is that the rotating scanning assembly of the ATMS is introducing a polarized contribution to the received signal, due to reflector emissivity. The flat-plate scanning reflector is a Beryllium plate with a thin layer of Zinc, and an outer layer of 0.6 micron Gold plating. It is well known that reflectors composed of a thin conductive layer can present much higher microwave emissivity than the theoretical values for bulk materials. This is likely due to the irregularities and granularity of the conductive layer. When viewed at a 45 incidence angle, as is the case for the ATMS configuration, this emissivity is polarized. As an example, if the emissivity were 0.37%, and for a physical temperature of 0 C, the resulting QV and QH contributions from the reflector emissivity would be as shown in the red curves of Figure 4. This is very near to the observed on-orbit biases, except that the on-orbit data shows a greater increase at large scan angles, especially at the cold calibration angles (81.7 85.0 ). No actual measurements have been made of the ATMS reflector emissivity, but experience from other programs indicate that the value hypothesized above is certainly plausible. For example, the original design of the SSMIS reflector (vapor-deposited Al on graphite epoxy) had about 1.0 % emissivity. A secondary factor that can affect the pitch-maneuver scan biases is the fact that the exo-atmosphere itself produces a small atmospheric emission, which makes a contribution that increases symmetrically with scan angle, regardless of polarization. This contribution was modeled as a quadratic function of scan angle, with 0.2 K brightness temperature at 90 degrees. The result of adding this to the reflector emissivity effect, as shown in Figure 4, is that the QV channels will have a somewhat greater magnitude of variation than the QH channels, and a notable increase at the cold calibration angles, which fits well with the on-orbit data. Another secondary effect is due to antenna sidelobe intercepts with the spacecraft, which occurs on the sun-side (i.e. negative scan angles), thus producing an asymmetric deviation from the sine-square and cosine-square functions. As seen in Figure 4, channel 1 has higher temperatures on the sun side (negative scan angles) than on the cold side. The same effect is seen for channels 2-15. For channels 16-22, the sun side has lower temperatures. This is probably because the antenna aperture for
channels 1-15 has a clear view to the horizon, with some intercept of the solar array. Channels 16-22, on the other hand, are obstructed by the CERES instrument, which may be reflecting radiation from cold space in the zenith direction. The observed asymmetric contributions are no more than about 0.2 Kelvin, which is consistent with the predicted worst-case spacecraft intercept contributions of up to 0.4 Kelvin. According to this proposed model for the scan-dependent bias, the magnitude of the effect will be proportional to reflector physical temperature. There will also be a scan-dependent reduction of reflectivity, introducing another error source, proportional to the scene brightness temperature. This means that correction algorithms to be applied operationally to scene data need to account for these additional factors. For example, Figure 5 shows example comparisons of the modeled scandependent bias that would occur when viewing earth scenes, either at the cold-case reflector temperature (as occurred during the pitch maneuver), or at the predicted hot-case reflector temperature. These plots show both cold and warm calibration positions (at 83 and 195 respectively) as well as the earth-view sector. The differences between these curves indicate there could be residual errors up to 0.5 K if the pitch-maneuver bias data were applied directly as a correction without modeling all the relevant parameters. It should also be noted that any correction used in the SDR algorithm should apply appropriate corrections to the warm and cold calibration views as well as to the scene viewing sector. Figure 5: Variability of predicted scan-dependent bias versus reflector temperature and scene temperatures SUMMARY Assessments of the on-orbit data from the NPP ATMS indicate all performance parameters are within expected values, confirming radiometric performance superior to AMSU. Furthermore, pitchmaneuver data has been used to develop a physical model for the scan-dependent bias effect, which has been a long-standing issue with cross-track scanning radiometers. Such a model can be used for developing a correction algorithm that could further reduce radiometric calibration errors relative to that of prior instruments.