934 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL /$ IEEE

Transcription

1 934 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 Analysis of the Special Sensor Microwave Imager/Sounder (SSMIS) Fields-of-View on DMSP F-16 David B. Kunkee, Senior Member, IEEE, Ye Hong, David A. Thompson, Michael F. Werner, and Gene A. Poe, Member, IEEE Abstract The Special Sensor Microwave Imager/Sounder (SSMIS) calibration and Earth scene fields-of-view (FOV) are examined using sensor diagnostic modes, a graphic simulation of the SSMIS and spacecraft vehicle, and an electromagnetic analysis of the SSMIS antenna. The on-orbit FOV for each calibration target was found to be larger than the sample locations utilized by the flight software for calibration suggesting that additional observing time for the each target is available. The optimum calibration sampling location appears to be slightly offset from the current sampling. However, it is not expected that this will result in degraded performance. The SSMIS Earth scene FOV was found to have noticeable edge-of-scan biases for several channels. The biases were examined and determined to be due to intrusions of the antenna feed FOVs of the main reflector antenna caused by the SSMIS calibration targets and their associated multilayer insulation (MLI) blanketing in the final flight configuration. A simple algorithm to address the edge-of-scan biases is applied and found to correct the biases to within K. Index Terms Calibration, microwave radiometry. BP BOS CSR DGS EO EO2 EOS FOV GRASP MLI PO TDRP WL NOMENCLATURE Beam Position. Beginning of Scan. Cold Sky Reflector. DMSP Graphic Simulator. Early Orbit. Early Orbit Mode #2 (A, B, C). End of Scan. Field of View. Generalized Reflector Analysis Software Package. Multilayer Insulation. Physical Optics. Temperature Data Record Processor. Warm Load. I. INTRODUCTION OPERATION of the Special Sensor Microwave Imager Sounder (SSMIS) on the F-16 Defense Meteorological Manuscript received March 15, 2007; revised August 14, D. B. Kunkee, Y. Hong, and M. F. Werner are with The Aerospace Corporation, Los Angeles, CA USA ( David.Kunkee@aero.org). D. A. Thompson was with The Aerospace Corporation, Los Angeles, CA USA. He is now with The Aerospace Corporation, Silver Spring, MD USA. G. A. Poe is with the Naval Research Laboratory, Monterey, CA USA. Digital Object Identifier /TGRS Satellite Program s (DMSP) Block 5D-3 spacecraft is very different from the previous DMSP 5D-2 mission series carrying the heritage DMSP Special Sensor Microwave Imager (SSM/I) [1] [3]. Although the SSM/I and SSMIS sensors are related with conical scan heritage, there are significant differences in the sensor operation, deployment and mounting, fields of view (FOV), and overall complexity. This is in addition to the defining characteristic of the SSMIS: its multifrequency atmospheric temperature and sounding channel suite. The SSM/I scans clockwise, whereas the SSMIS scans counterclockwise as viewed from the top. The SSM/I deploys in a one-step process, but the SSMIS deployment scheme requires translation and rotation of the sensor assembly to set the SSMIS further away from the 5D-3 spacecraft, which in turn allows a greater active scan sector and creates an improved Earth scene FOV. The SSM/I utilizes a single multiband feedhorn located at the prime focus of its 0.6-m offset parabolic reflector, while the SSMIS utilizes six separate feedhorns positioned along the azimuthal focal line. The SSMIS design has a unique FOV, geolocation, and along-scan averaging for each feedhorn. Because of the additional complexity due to multiple feedhorns and polarizations, a detailed analysis of the sensor FOV and potential alongscan biases was necessary. In normal operations, radiometric scene data are collected only over the predetermined active scan range and averaged onboard according to a scheme that takes into account the inherent spatial resolution and noise performance of each group of channels [2]. Data-rate limitations imposed by the spacecraft do not allow full downlink of continuous data from all 24 channels. To characterize the on-orbit sensor operation more fully, radiometric data from the SSMIS Early Orbit (EO) mode collections in October 2003 and January 2005 were used to evaluate the sensor FOVs and identify potential along-scan biases. The SSMIS EO modes collect radiometric data at full spatial resolution (no onboard averaging) over the full 360 scan, enabling evaluation of: 1) the beginning and the end of Earth scene FOVs and 2) alignment of calibration observations with the calibration targets. Data from the EO and normal modes have indicated biases near the edges of the active Earth scene FOV; however, the calibration FOVs were found to be uniform with the possibility of expanding the observation time for both the hot and cold calibration observations. To support analysis of the EO data, a detailed graphic simulation was developed to visualize the relationship of the /$ IEEE

2 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F Fig. 1. SSMIS mounted on the DMSP Block 5D-3 spacecraft in a view generated by the DMSP Graphical Simulator (DGS). The SSMIS is shown in beam position 200 near the end of the active Earth scan field of view. The SSMIS scans counterclockwise when viewed from the top. individual antenna feed FOVs with respect to the calibration load assembly and spacecraft. The calibration load assembly is composed of a warm load and a cold sky reflector. The SSMIS EO modes and graphic simulator are described in Section II. EO data collection, processing, and analysis are described in Section III. In Section IV the warm load (WL) FOV analysis is presented, and Section V describes the cold sky reflector (CSR) FOV. The Earth scene FOV is characterized in Section VI, and analysis of the FOV intrusions follows in Section VII by applying the General Reflector Antenna Software Package (GRASP) version 9 [4]. The GRASP program is an electromagnetic analysis tool that uses physical optics (PO) to evaluate electromagnetic characteristics of the antenna subsystem. A correction scheme for the FOV intrusions is then defined and applied. Section VIII summarizes the principal results of the SSMIS F-16 FOV analysis. To aid the reader a list of commonly used nomenclature appears at the beginning of the paper. II. DMSP GRAPHIC SIMULATOR SSMIS MODEL DEVELOPMENT The DMSP F-16 SSMIS sensor, as mounted on the Lockheed Block 5D-3 spacecraft, is shown graphically in Fig. 1 using the DMSP Graphic Simulator (DGS). Note that the F-16 spacecraft velocity vector points in the Y -direction, 1 and X is the nadir direction. The DGS tool was originally developed using an open source graphics library (OpenGL) with simplified computeraided design (CAD) models of the DMSP spacecraft and sensors. The model is used to visualize the spacecraft environment with respect to the Earth, sun, and moon to aid analysis of scan coverage, solar and lunar illumination, FOV intrusions from 1 The 5D-3 series can fly in the Y -or+y -directions depending on whether the prescribed orbit has an ascending nodal crossing time before or afternoon. The sun must stay on the Z side of the spacecraft. For F-16, the direction of travel is Y, causing the Earth scene FOV to be in the forward direction. Fig. 2. Top view of the SSMIS mounted on the DMSP F-16 spacecraft (main reflector not shown) showing the relative position of its 6 antenna feeds (in sequence beginning at the top left): Ka-band, V-band for Lower Atmospheric Sounding (LAS), G-band, W-band, V-band for Upper Atmospheric Sounding (UAS), and K-band. The cold sky reflector and warm load remain stationary with respect to the spacecraft. the spacecraft or the sensor; perform near-term coverage and revisit times; and determine viewing windows for satellite data downlinking. In support of the SSMIS FOV analysis, many new capabilities were added to the DGS simulation, including a basic geometrical optics (GO) visualization of each channel s antenna pattern, and FOV. The relative position of each SSMIS antenna feed can be seen in the top view of SSMIS as mounted on the DMSP 5D-3 spacecraft (Fig. 2). The SSMIS spins counterclockwise when viewed from the top so that the warm calibration target is viewed by the feeds prior to the cold target. The 360 scan is divided into 450 beam positions (BPs) of 0.8 each. One BP is the minimum sampling interval of the SSMIS, representing a 4.22-ms integration period. The Ka-band antenna feed views the calibration targets first, followed by the V-band lower-air sounding (LAS) feed. The last antenna feed to view the calibration targets in each rotation is the K-band. In contrast to the sequence of calibration observations, the sequence for the Earth scene FOV is reversed. The K-band antenna feed (Channels 12 14) is the first in the sequence of antenna beams to sweep through the Earth scene, whereas the last in the sequence are the Ka-band channels (15 and 16). The SSMIS FOVs for each antenna feed are defined in Table I and show the range of the Earth scene s beginning and end and the warm and cold calibration observations. The DGS model provides visualization of these sequences allowing the user to scan the SSMIS while the model graphically shows illumination of the main reflector from the antenna feeds as well as the main antenna pattern. In this manner, potential blockages of the antenna beam from adjacent structures can be identified. For reference, the complete SSMIS channel set and their characteristics can be found in [2]. Flexibility to optimize the location of calibration samples can be achieved with changes to the SSMIS flight and ground processing software.

3 936 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 TABLE I AZIMUTHAL LOCATION (BEAM POSITION) OF SSMIS CALIBRATION AND EARTH SCENE FOVS FOREACH ANTENNA FEED To verify proper alignment of the calibration and Earth scene FOVs as determined prelaunch, an early orbit (EO) mode data collection was performed soon after the instrument began operating. The SSMIS flight software configuration provides five different modes of operation that can be commanded from the ground. Of these modes, there are four EO modes each with unique attributes tailored to aid the instrument check-out, integration and test, on-orbit check out, and normal operation. Two EO modes were used in the initial FOV analysis, EO2B and EO2C. These modes provide radiometric counts from Channels 1, 6, and 8 (in EO2B) and Channels 12, 15, and 18 (in EO2C) over the entire 360 azimuthal scan. A third mode, Early Orbit 2A (EO2A) provides radiometric counts from all 24 SSMIS channels; however, the data are separated into seven independent scan segments of 60 BP each and one 30 BP segment to cover 450 BP overall. This means that each scan segment is repeated after every eight scans. III. SSMIS EO DATA PROCESSING AND ANALYSIS The F-16 SSMIS was placed in EO2A, B, and C modes shortly after spin-up to confirm proper operation and to evaluate the sensor FOVs. The data collection period in October 2003 included two orbits of EO2A data, six orbits of EO2B data, and two orbits of EO2C data. A. Calibration FOVs The EO2B and EO2C data were first used to view and verify WL and CSR calibration observations for all antenna feeds. The timing information in Table I indicates that the average radiometric counts from BP and are used as the WL and CSR calibration observations, respectively, for the K-band antenna feed. Similarly, BP and are utilized for the WL and CSR calibration observations for the G-band antenna feed. Accordingly, EO2 mode radiometric counts from the K-band (lowest SSMIS channel frequency) and the G-band (highest SSMIS channels frequency group) are shown in Fig. 3 for the WL and Fig. 4 for the CSR. In these figures, BPs used in normal mode to collect radiometric calibration data from the WL are highlighted in gray and appear to have the largest and uniform values validating the WL BP assignments. Similarly, BPs used in normal mode to collect radiometric data for the cold calibration are also highlighted in gray and uniformly show the lowest counts validating the SSMIS Cold Calibration BP assignments. Note the zigzag response between odd and even beam positions; Fig. 3. Radiometric counts from SSMIS mode EO2C for (a) Channel 12 and EO2B for (b) Channel 8 showing the sensor response in radiometric counts over the warm load position. The zig-zag response is caused by small differences between the analog integrator circuits, one used for odd numbered beam positions and the other used for the even numbered positions. The shaded area represents the beam positions used for sensor calibration. the odd and even beam positions for each channel use unique integrator circuits in the SSMIS. This leads to slightly different radiometric counts for the same scene brightness. 2 2 The SSMIS employs two integrator circuits to provide continuous observation of the Earth scene. While one integrator is read, the other continues to collect scene data. In normal mode, the relative gain and offset of the two integrator circuits is removed by a calibration routine.

4 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F (see Table III). The terms H1 H4 indicate the four BPs used for the WL observation, C1 C4 are the four BPs used for the cold sky observation and N is the total number of scans. This calibration approach allows a radiometric brightness temperature, ˆT P to be determined from the raw EO radiometric counts. Three additional steps of EO2 data processing were also performed to aid the FOV analyses. First, the radiometric temperature at odd and even BPs was averaged to eliminate the bimodal response ˆT P (i, j +0.5) = ˆT P (i, j)+ ˆT P (i, j +1). (2) 2 Fig. 4. Radiometric counts SSMIS mode EO2C for Channel 12 (a) and Channel 8 (b) in their respective beam positions for measurement of the cold calibration scene. B. Calibration of EO2 Data Radiometric counts from EO2 mode data were calibrated using average counts from WL and CSR observations for FOV analysis purposes. The shaded region identifies the four BPs that were averaged to obtain the calibration target value in radiometric counts. A fixed calibration based on the average radiometric counts from the four WL and four CSR calibration observations averaged for all of the available scans was performed according to (1) over the entire series of orbits within the EO2 mode, ˆT (i, j) P = 1 4N N n=1 {[ H4 m=h1 C(n, m) P C4 m=c1 C(n, m) P ] 1 } [T (H) T (C P )] C(i, j) P (1) where ˆT (i, j) P is the estimated brightness temperature for channel P for scan number i and BP = j, C(n, m) P is the raw count for scan number n and BP = m, T (H) is the average measured hot calibration load temperature, and T (C P ) is the effective cold space temperature for the Channel P Second, to overlay responses of the discrete antenna feeds from a fixed scene location (such as a calibration target scene), data from each channel must be offset first by an integer number of BPs and then by a fractional BP representing the precision integrator, sampling, and antenna pattern offsets for each channel with respect to the nearest integer BP as shown in Table II. For example, Channel 18 WL samples occur at BP , and Channel 8 WL samples at BP as indicated in Table I. However, Channel 18 is sampled 0.2 (0.250 BP) earlier than Channel 8 for the same integer-valued BP. The integer BP values are associated with the absolute angular position of the SSMIS canister. The integer BP and fractional BP absolute timing offsets found in Table II were incorporated into in the EO2 data sets to allow comparison of the calibration FOVs. Finally, data from each feedhorn must be overlaid by shifting, or offsetting, BPs to a common standard to show common FOVs. For the remaining analyses, the K-band BPs are adopted as the standard. For example, when K-band (e.g., Channel 12) data are shown with Ka-band (e.g. Channel 15) data, the beginning of scan for both channels will appear as BP 48, the beginning of scan for K-band according to Table I, even though the beginning of scan for Channel 15 is actually BP 70. IV. WARM LOAD FOV Calibrated data from the EO2B and EO2C mode data collection were overlaid to show the detailed WL sampling, Fig. 5. The large-scale Fig. 5(a) shows the relationship of the selected WL sampling BPs with respect to the stable region of brightness temperatures provide by the warm load. The smaller scale Fig. 5(b) shows the interaction of each channel over the center of the WL region. Each channel was calibrated according to (1) at the four BP samples indicated by the shaded region (all channels mapped to K-band standard BPs) to an effective WL temperature of K for all channels. This is the average value measured by three Platinum Resistance Thermometers (PRTs) located on the back of the WL each with an absolute accuracy of 0.1 K. For Channel 15, many of the Ka-band BPs are not shown in Fig. 5 due to a break in data continuity within the EO2C mode at the position near the WL observation. For the data shown from Channel 15, the WL calibration points used in (1) were moved to equivalent BP Fig. 5(b) shows the fine-scale characteristics of the on-orbit WL view. First, it appears that by delaying the WL calibration

5 938 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 TABLE II INTEGER AND PRECISION SAMPLING OFFSETS FOR EACH ANTENNA FEED FOR CALIBRATION AND EARTH SCENE because the G-band antenna feed contour on the SSMIS WL is on the order of the individual tine dimensions and hence provides the strongest signature magnitude. Other feed contours cover several tines blurring the effect. Additional factors that may influence the fine-scale response are solar illumination of the WL tines [5], [6] and possible standing waves between the antenna feed and warm load structure. The individual SSMIS radiometers may respond to the instantaneous changes in effective distance between the WL and the antenna feed as the tine structures pass in front of the high-frequency feeds [7]. For the channels shown in Fig. 5, these effects may also be the strongest for Channel 8 because the high-frequency channels do not include an isolator between the antenna feed and receivers. Fig. 5. Detail of on-orbit Warm Load FOVs as viewed in EO2B, Channels 1, 6, and 8, and EO2C Channels, 12, 15, and 18: (a) warm load observations calibrated to K, (b) small-scale variations in measured brightness temperature across the WL FOV. Values computed using (1). The shaded region indicates BPs used for calibration in Normal Mode. sampling by 3 or 4 BP would optimize the WL response for all channels. However, the difference in effective WL T B would be minimal, e.g. < 0.05 K and likely result in no substantive improvement to the sensor calibration. Second, there is an oscillatory behavior in the response of nearly all channels to the WL as a function of BP. In general, the phase of the signatures in Fig. 5(b) depends on the convolution of the near-field antenna feed response and the WL temperature distribution to determine if the WL effective temperature at the tine peak or valley dominates the response at any specific BP. For the channels included in Fig. 5, this will impact Channel 8 most significantly V. C OLD SKY FOV The CSR FOV is shown in Fig. 6 using (a) large- and (b) small-scale brightness temperature ranges to illustrate the cold sky scene variability as measured over the entire CSR on a fine scale. The shaded region on the left represents the F-16 calibration sample location, and the shaded region on the right represents the same samples 10 BP later, which is the optimal sampling location considering all channels excluding Channel 8. Although the F-16 CSR calibration sampling was not changed, the sampling location was changed for the DMSP F-17 SSMIS accordingly. Two characteristics in Fig. 6 are readily apparent. First, the cold sky temperatures are slightly different for every channel as a result of applying unique values for the cold calibration temperatures from Table III to (1) for the EO2 mode calibration analysis. The effective cold sky temperatures in Table III are used in normal mode because the Rayleigh-Jeans approximation of Planck s blackbody function begins to have appreciable error when brightness temperatures near zero are used in the higher millimeter-wave frequency region. Hence, the effective cold sky temperature is adjusted upward so that a normal twopoint calibration can be used with minimal residual error in expected range of Earth scene brightness temperatures. Second, there is an apparent trend in Channels 8 and 18, the two highest frequency channels shown in Fig. 6, showing the brightness temperatures to be higher at the beginning of the CSR FOV compared to the end by approximately K. This is based on the relatively symmetric response of the lower frequency channels. It is not certain if this effect is due to geometry of the reflector or possible self-emission from the reflector. The DGS simulation indicates that the sun illuminates the front

6 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F calibration observations for each scan. Additional samples may be particularly important for the narrow band channels that have a high inherent NEDT associated with each 4.22 ms integration sample [8]. From Fig. 5, it appears that at least 8 BPs could be used for the WL calibration observation; the K-band data in Fig. 6 suggests that 12 BPs of the CSR view are uniform to within 0.1 K. Fig. 6. Detail of on-orbit Cold Sky Reflector FOVs as viewed in EO2B, Channels 1, 6, and 8, and EO2C Channels, 12, 15, and 18: (a) CSR FOV with calibration to the frequency-adjusted cold space values in Table III, (b) smallscale plot of the brightness temperature variability across the CSR FOV. Values computed using (1). The shaded vertical bar on the left indicates the location used for calibration on F16. The calibration location has been moved to the location shown by the shaded bar on the right for later SSMIS flights. TABLE III COLD SKY CALIBRATION TEMPERATURES USED IN NORMAL MODE AND EO2 DATA ANALYSIS surface of the CSR during a small portion of the orbit; however, these occurrences do not correlate well with trends in the cold sky counts. Further discussion about the solar illumination and SSMIS reflector emission can be found in [6]. Overall, the calibration FOVs shown in Figs. 5 and 6 indicate that additional WL and CSR samples may be available to reduce the noise equivalent difference temperature (NEDT) of VI. EARTH SCENE FOVS The Earth scene FOVs were examined by averaging EO2B and EO2C data from multiple orbits for each individual BP over the active scan region of the Earth scene. Additional BPs beyond the edge of the Earth scene were also included to better understand the trend of the data at the edge of the scan. The data were not filtered based on the type of scene, and all scenes were included in the EO data analysis. Radiometric data collected within the Earth scene, if averaged over an extended period of time (many orbits), should trend to a uniform value over the entire swath if each portion of the scan has the same scene statistics, [9] and no FOV intrusions exist that could bias the data in specific areas of the scan. Variations within the Earth scene region for this EO2 analysis can be placed into three categories: 1) residual scene variability that could not be eliminated by sufficient duration of averaging; 2) systematic variations of the scene due to observing geometry and orbital characteristics that will not be reduced by additional averaging; and 3) systematic biases introduced by the scan FOV at the edges of the active scan. Typically, 3) is due to the antenna main beam sidelobes viewing the spacecraft or the parts of the CSR or WL near the edges of the swath. Beam positions that compose the Earth scene FOV are defined in Table I for each antenna feed. Similar to the calibration FOV analysis, EO2B and EO2C data containing radiometric data from Channels 1, 6, 8, 12, 15, and 18 are overlaid together on the same graph to show the interrelationship between SSMIS antenna feed characteristics. The Earth scene BP offsets for this analysis are found in Table II. As previously stated, Channel 12 (K-band antenna feed) is used as the reference; however, the offsets for all channels are now positive (later) because Channel 12 begins its Earth scene FOV first at BP 48. For the Earth scene FOV analysis, radiometric calibration is performed according to (1) using the values in Table III, however, the average brightness temperatures for each BP are offset to a common value for all channels to enable data overlay. The initial set of EO2C data was limited to two orbits of data and was of insufficient quantity to establish brightness temperature characteristics across the active Earth scene with reasonable uncertainty. The EO2C Channel set includes Channels 12, 15, and 18, all having significant variability due to surface type and clouds. A longer collection period (six orbits) occurred for EO2B mode (Channels 1, 6, and 8); however, the along-scan characteristics of the data were still uncertain below 1 K. Detailed comparisons between EO2B and EO2C with multiple three-month averages of normal mode data were successful for establishing the along-scan characteristics near the edge of the scan. Both sets of data, EO2 and Normal Mode, indicated consistent edge-of-scan biases in Channels 1, 12, and 15.

7 940 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 Fig. 8. (a) DGS simulation of the SSMIS in BP 48 at the BOS of the Channel 12 (K-band antenna feed) Earth scene FOV. The reflector has been removed to show the orientation of the antenna with respect to the cold sky reflector. In (b) the K-band FOV to the main reflector edge is illustrated by the cone. The FOV intrusion is noted by the arrow. VII. ANALYSIS OF FOV INTRUSIONS During the course of the SSMIS calibration and validation, many analysis tools were developed to facilitate understanding of the sensor operation and to improve radiometric and environmental data quality. The DGS as described in Section II is an important tool for providing insight on the SSMIS operation. Aspects of the SSMIS design responsible for of the edge of scan biases suggested by Fig. 7(a) and (b) may be identified using the DGS simulation to visualize operation of the SSMIS and spacecraft at specific beam positions. Fig. 7. Summary of early orbit (EO) data analysis from EO2B and EO2C modes for evaluation of the SSMIS Earth Scene FOVs. Data from all channels have been co-aligned using Channel 12 BPs. The BOS characteristics (a) show edge-of-scan biases in Channel 12. The EOS characteristics show edge-of-scan biases in Channel 1 and Channel 15 data. The value of EO2 data analysis is in the ability to view the sensor response beyond the edges of the Earth FOV to help understand the data trends and identify the root cause of any adverse characteristics attributable to the sensor or spacecraft operation. Fig. 7 shows the edge-of-scan characteristics for Channels 1, 12, and 15 adjusted (offset) to 150-K relative brightness temperature for comparison purposes. Channel 12 brightness temperatures show an increasing trend from the beginning of scan at BP 48 until BP 55, Fig. 7(a). In this example, it is clear that BPs near the beginning of the scan (BOS) are biased low. All other channels at the BOS appear to have no along-scan bias within this portion of the scan sector. However, Channels 1 and 15 indicate a systematic decrease at the end of the scan (EOS) as shown in Fig. 7(b). The edgeof-scan characteristics elucidated by Fig. 7 require additional investigation to determine the root cause and validate an approach for correction of the biases. The center portion of the SSMIS scan was examined during the course of the EO2 data collection and early normal mode collection and did not appear to have any significant anomalies. A. Graphic Simulation Using DGS The DGS model was used to display simplified antenna feed patterns illuminating the main reflector to visualize the SSMIS Earth scene FOV geometry between the antenna feeds and the main reflector during the active scan (Earth scene FOV). At the BOS, the Channel 12 antenna feed pattern was simulated to investigate the edge-of-scan bias shown in Fig. 7(a). Fig. 8 illustrates the SSMIS in BP 48 at the BOS with the main reflector removed. This allows a simple K-band antenna feed pattern, as depicted by the cone, to be examined. In this BP, the cone extends from the antenna feed to the outside edge of the main reflector and intersects the CSR support structure in the area depicted by the arrow. This FOV intrusion between the antenna feed and SSMIS main reflector is a possible explanation for the BOS bias. For reference, the main reflector illumination edge taper for K-band is 20 db. Similarly, with the SSMIS positioned near the Ka-band EOS, the antenna feed pattern impinges on the top of the WL shroud as shown in Fig. 9. Accordingly, Figs. 8 and 9 provide a possible root cause, based on geometrical optics, for the edge-of-scan biases observed in the data of Channels 12 and 15 (Fig. 7). Although the FOV intrusions illustrated in Figs. 8 and 9 are consistent with observed T B data, the actual antenna patterns are much more complex than this simple representation, based on the geometrical optics [10]. Therefore, an electromagnetic analysis is required to evaluate each of the potential FOV intrusions identified by DGS.

8 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F Fig. 9. (a) DGS simulation of the SSMIS in BP 250 at the EOS for Channel 15 (Ka-band antenna feed) Earth scene FOV. In (b) the top of the warm load shroud intrudes into the FOV cone from the Ka-band feed as noted by the arrow. B. Analysis of FOV Intrusions Using GRASP For the following analysis, we apply the GRASP-9 to model the potential FOV intrusions identified in Figs. 8 and 9. The GRASP program uses PO analysis techniques to provide an accurate electromagnetic analysis of the SSMIS antenna and calibration assembly. For this analysis, the main reflector geometry of the SSMIS was available and incorporated into GRASP; however, detailed design data of the individual antenna feeds were not available to create a physical model of the antenna feeds. This uncertainty was overcome in the GRASP simulations by matching measured antenna patterns from the SSMIS main reflector and individual antenna feeds to define the characteristics of the ideal antenna feeds used in GRASP. Within GRASP, critical areas of the SSMIS WL, CSR, and the edge of the spacecraft were modeled as perfect electric conductors (PEC) in a simplified geometry. In this manner, the FOV of each antenna feed to the main reflector antenna was evaluated. During this process of defining the simplified CLA geometry at the level of detail required by GRASP, it was found to be necessary to account for multilayer insulation (MLI) blankets covering the back of the CSR and top of the WL shroud. The MLI blankets were added prior to launch for additional thermal stabilization of the WL and CSR structures. Accordingly, geometry of the blanketed CSR and WL was developed for GRASP based on photographs of the SSMIS flight unit and later refined by measuring the launch-ready F18 SSMIS flight unit. Biases attributable to the WL shroud, CSR mounting structure, and spacecraft body were computed by configuring GRASP to calculate the fraction of total power from the antenna beam impinging on each of the intrusions. The beam fraction was then used to calculate the scene bias using the following equation: ˆT (φ) =P i (φ)(t CB T Center ) (3) where ˆT is the estimated impact of the FOV beam intrusion in K as a function of scan angle, φ, T Center is the mean brightness temperature at the center of scan, T CB is the cosmic background brightness temperature, and P i is the beam power fraction impinging on item i, calculated by GRASP, where i is the CSR, WL, or the spacecraft. Fig. 10. GRASP analysis of Channel 12 beginning-of-scan FOV intrusion from the CSR mounting and MLI thermal blankets. SSMIS Channel 12 data from the calibrated EO2C mode data collection is provided for comparison. The modeled response due to simulated FOV intrusions in GRASP from the CSR assembly near the beginning of the K-band Earth scene FOV (BP 48) is shown in Fig. 10. The EO2 data from Channel 12 is normalized to a value near the center of scan, K. Values of ˆT derived from (3) are shown overlaid with Channel 12 EO2 mode data appearing in Fig. 7(a). Values of P SC from GRASP were generated with 1) the single CSR mounting panel and 2) the CSR mount and the simulated MLI covering. As indicated, calculations that include FOV intrusion from the MLI blankets provide better agreement with EO2 data at the BOS. In a similar analysis at the EOS, the modeled response of Channels 1 and 15 from the WL shroud and MLI blanket FOV intrusions are shown in Fig. 11(a) and (b). The Channel 1 data are from EO2B mode collection also shown in Fig. 7 and normalized to K. Channel 15 data are again normalized to K, its value near the center of scan. Note that the fraction of total beam power blocked by the WL and its MLI blanket is much less for Channel 1 because the overall impact at the EOS is 2 K, and the average T B is 246 K. If T Center T CB, and if (3) is valid, or equivalently, the apparent emissivity of the MLI, ε MLI 0, 3 then ˆT (φ) T B = P WL(φ)(T CB T Center ) T B P WL (φ). (4) Therefore the amount of power from Channel 1 is P WL at BP 245 (EOS) compared to P WL 0.04 for Channel 15 at BP 250. The GRASP simulation as configured for this study is nearly polarization independent. The analysis computes the power of the antenna feed beam pattern incident on the CSR and WL 3 The MLI utilized on SSMIS contains metalized layers of material. A similar MLI from a different manufacturer was found to have an RF conductivity of 0.7 MS/m near 10 GHz [11].

9 942 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 TABLE IV RELATIVE FOV EDGE-OF-SCAN BIASES FOR DUAL-POLARIZED SSMIS ANTENNA FEEDS BASED ON NORMAL MODE DATA COLLECTED FROM F-16 ORBITS Fig. 11. GRASP analysis of Channel 1 (a) and Channel 15 (b) showing the end-of-scan FOV intrusion from the WL shroud and associated MLI thermal blankets. SSMIS Channel 1 and Channel 15 data from calibrated EO2B and EO2C mode data collections are provided for comparison. Fig. (b) also shows the GRASP simulation of the FOV impact from the spacecraft edge only. intrusions. The value of T is, however, highly polarizationdependent due to different T B for every channel. Two examples are provided in Table IV using normal mode data averaged for approximately one week (F-16 orbits ). The example indicates identical P WL or P CSR for both polarizations based on the observed edge-of-scan biases. These examples using normal mode data show consistency with the GRASP simulation approach with a data set independent of EO2. Also included in Fig. 11(b) are GRASP results from power incident on the WL cover wing, the lower horizontal portion of the WL cover. It is seen that the WL wing does not impact Channel 15 FOV until BP 256, which is well past its Earth scene FOV. Similarly, the spacecraft body does not impact Channel 15 FOV until BP 260, also well past the Earth scene FOV for that channel. Although it is not shown, a similar result was found for Channel 1; the spacecraft did not impact Earth FOV of Channel 1 until BP 252, whereas the EOS is at BP 245. Fig. 12. LAS Channel group EOS biases. (a) All channels have similar trend and biases near the EOS. (b) Detailed comparison of relative channel end-ofscan biases suggests slight dependence on antenna feed pattern as a function of frequency. Fig. 12 shows that T B at the EOS for the LAS channels changes from 2 K (Channel 1) to 1 K for Channel 5. The difference in end-of-scan bias may be explained by variability of the antenna feed pattern as a function of frequency. Antenna feed patterns supplied by the SSMIS manufacturer suggest that the beam taper at the main reflector edge increases 3 dbor more for Channel 5 compared to Channel 1. Measured farfield patterns of Channel 1 were matched to simulated farfield patterns generated by GRASP to determine the correct illumination taper for SSMIS F-16. However, far-field main antenna patterns were not available for Channel 5, so a sensitivity analysis was performed by increasing the illumination taper for the Channel 5 FOV simulation by 3 db. The impact of this

10 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F Fig. 13. GRASP analysis of Channel 1 and Channel 5 showing the impact to FOV intrusions as a function of illumination taper of the antenna feeds. TABLE V MAXIMUM ANTENNA BEAM POWER FRACTION INCIDENT ON THE BLANKETED CSR OR WL FOR ALL SSMIS ANTENNA FEEDS AND NORMALIZED T B BASED ON T B = 250 K AND T CB AS LISTED change on ˆT was found to be consistent with LAS EO data shown in Fig. 12 when the data were normalized to the observed T B for each channel, T B (1) K, T B (5) K (Fig. 13). Finally, using GRASP, the incident power on the CSR mount and WL shroud (with blankets) was computed for the full Earth scene FOV including additional BP before and after the BOS and EOS, respectively. The impact, ˆT (φ) was computed based on a standard T B = 250 K, T CB, and ε MLI 0 for all channels to determine the maximum potential bias over the Earth scene scan. Values of ˆT (φ) related to each antenna feed are shown in Fig. 14 and the values at the edge of the scan listed in Table V. Biases estimated for the UAS, W-, and G- band antenna feeds have not been observed in the EO or normal mode data. This may attest to the uncertainty still remaining in the GRASP FOV modeling, particularly at the BOS. All scan-dependent biases identified by averaging the normal mode for extended periods were corrected in the temperature data record processor (TDRP) by determining a constant, L(φ), similar to 1 P (φ) but derived from a large ensemble of normal mode data. To the first order T A (φ) =L(φ)T Scene +[1 L(φ)] T CB (5) Fig. 14. GRASP analysis of the K-band, UAS, W-band, G-band, LAS, Ka-band antenna feeds at the beginning (a) and end of scan (b) normalized to 250 K with emissivity of FOV intrusions assumed to be very low, (ε MLI 0). where T A is the antenna temperature, T Scene is the scene temperature, and T CB is the cosmic background brightness temperature. Because 1 L 1 T Scene T A(φ) L(φ) L(φ) = T A (φ) T A (φ : Center) this approach was applied to SSMIS TDR data and is believed to correct scan-dependent biases to within K over the sensor swath. 4 4 The SSMIS Doppler compensation state may impact calibration of the lower and upper air sounding channels (1 7, and 19 24). Accordingly, the Doppler compensation state must remain the same during the data collection supporting (7) and application of (6) to correctly address the scene FOV biases. (6) (7)

11 944 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 46, NO. 4, APRIL 2008 VIII. CONCLUSION In this paper we have attempted to characterize the SSMIS calibration and Earth scene fields of view using a combination of 1) data collected from the SSMIS operating in the EO2 diagnostic modes, 2) a graphical simulation of the SSMIS on the DMSP spacecraft, and 3) the GRASP antenna analysis package. It was found that both the warm and cold calibration target FOVs could be expanded by approximately four to eight additional beam positions which would provide in principle, reduced noise in the calibration observations. The Earth scene FOV for the K-band, LAS, and Ka-band antenna feeds displayed edge-of-scan effects that were attributed FOV intrusions from the CSR mounting support structure, the warm load cover (shroud), and the MLI thermal blanketing required for thermal control in flight. The size and geometry of the FOV intrusions were estimated using photographs of a similar SSMIS flight unit and reaffirmed by measuring a similar flight unit. By modeling the power from the antenna feeds illuminating the FOV intrusions, an estimate of the impact in T A was found to be generally consistent with measured data from the SSMIS. The FOV analysis was run for all antenna feeds to investigate the potential of additional FOV intrusions that either may not have been noticed or are incorrectly attributed. One reason this might occur is that reflective characteristics of the MLI may be significantly different above 60 GHz, suggesting that FOV intrusions in the higher frequency channels may lead to biases in the scene that are dependent on the temperature of the MLI surface. Analysis using GRASP, for the W-band, G-band, and UAS antenna feeds suggested only very minor impacts. Corrections to scan-position-dependent biases were based on the amount of FOV intrusion at each beam position and determined by averaging several months of data and scaling the observed brightness temperatures to create a uniform average scene temperature over the swath. GRASP analysis confirms the validity of this approach by showing that scattering from cold space into the antenna feed from the WL and CSR structures and their associated thermal blanketing is consistent with biases observed in the radiometric data. ACKNOWLEDGMENT The authors would like to thank Donald Boucher of The Aerospace Corporation for supporting the initial development of the DGS model. The efforts of the SSMIS Cal/Val team were performed under support from the USAF DMSP program. SSMIS data were provided by the Fleet Numerical Meteorology and Oceanography Center and the Air Force Weather Agency. REFERENCES [1] J. P. Hollinger, J. L. Peirce, and G. A. Poe, SSM/I instrument evaluation, IEEE Trans. Geosci. Remote Sens., vol. 28, no. 5, pp , Sep [2] D. B. Kunkee, G. A. Poe, D. J. Boucher, S. D. Swadley, Y. Hong, J. E. Wessel, and E. A. Uliana, Design and evaluation of the first Special Sensor Microwave Imager/Sounder, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp , Apr [3] M. Colton and G. A. Poe, Intersensor calibration of DMSP SSM/I s F-8 F-14, , IEEE Trans. Geosci. Remote Sens., vol. 37, no. 1, pp , Jun [4] General Reflector Antenna Software Package, TICRA, Copenhagen, Denmark. Version 9 Reference Manual. [5] E. M. Twarog, W. E. Purdy, P. W. Gaiser, K. H. Cheung, and B. Kelm, WindSat on-orbit warm load calibration, IEEE Trans. Geosci. Remote Sens., vol. 44, no. 3, pp , Mar [6] D. B. Kunkee, S. D. Swadley, G. A. Poe, Y. Hong, and M. F. Werner, Special Sensor Microwave Imager Sounder (SSMIS) radiometric calibration anomalies Part I: Identification and characterization, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp , Apr [7] D. M. Jackson, Calibration of millimeter-wave radiometers with application to clear-air remote sensing of the atmosphere, Ph.D. dissertation, Georgia Inst. Technol., Atlanta, GA, Jul [8] S. D. Swadley, G. A. Poe, W. Bell, Y. Hong, D. B. Kunkee, I. S. McDermid, and T. Leblanc, Analysis and characterization of the SSMIS upper atmospheric sounding channel measurements, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp , Apr [9] F. J. Wentz, P. Ashcroft, and C. Gentemann, Post-launch calibration of the TRMM Microwave Imager, IEEE Trans. Geosci. Remote Sens., vol. 39, no. 2, pp , Feb [10] R. C. Hanson, Microwave Scanning Antennas. New York: Academic, [11] A. Prata, Personal Communication. David B. Kunkee (S 88 M 96 SM 04) received the Ph.D. degree in electrical engineering from Georgia Institute of Technology, Atlanta, in He joined The Aerospace Corporation, Los Angeles, CA, in 1995 and is currently an Associate Director. He is a Technical Advisor and a Sensor Scientist with the National Polar-orbiting Operational Environmental Satellite System (NPOESS) integrated program office involved with the development and planning for the new NPOESS Microwave Imager/Sounder. He is also a core member of the Defense Meteorological Satellite Program s Special Sensor Microwave Imager/ Sounder Calibration/Validation team. Dr. Kunkee is a member of the International Union of Radio Science, URSI (Commission F) and has served on the National Academies Committee on Radio Frequencies. He is currently Editor of the IEEE GEOSCIENCE AND REMOTE SENSING SOCIETY (GRS-S) NEWSLETTER and the past Chair of the GRS-S Technical Committee on Frequency Allocations in Remote Sensing. Ye Hong received the B.S. degree in electrical engineering from Zhejiang University, Hangzhou, China, the M.S. degree in electrical engineering from the Chinese Academia of Sciences, Beijing, China, and the Ph.D. degree in satellite meteorology from Texas A&M University, College Station. She is currently with The Aerospace Corporation, Los Angeles, CA, working on the Defense Meteorological Satellite Program SSMIS Cal/Val and supporting the NPOESS MIS algorithm development. Prior to joining Aerospace, she was a Research Scientist and worked on precipitation algorithm development and data analysis for the NASA TRMM project, as well as the TRMM Science Data and Information System. Dr. Hong is a member of the AMS and AGU. David A. Thompson received the B.S.E.E., M.S.E., and Ph.D. degrees from the University of Texas, Austin, in 1994, 1998, and 2002, respectively. In 2002 he joined the Antenna Systems Department at The Aerospace Corporation and became involved with characterization and analysis of the Defense Meteorological Satellite Program Special Sensor Microwave Imager/Sounder. In 2008 he joined NPOESS Space Systems Department of The Aerospace Corporation in Silver Spring, MD, as a Project Engineer. Dr. Thompson is a member of the Antenna Measurement Techniques Association (AMTA).

12 KUNKEE et al.: ANALYSIS OF SSMIS FIELDS-OF-VIEW ON DMSP F Michael F. Werner received the B.A. degree in astrophysics from Williams College, Williamstown, MA, in 1976, and the M.S. degree in aerospace science from the University of Michigan, Ann Arbor, in Since 1980, he has been with The Aerospace Corporation, El Segundo, CA, where he is currently a Senior Engineering Specialist in the Flight Mechanics Department. His concentration has been in engineering visualization and simulation in order to evaluate the design and performance of satellites, launch vehicles, experimental aircraft, and related sensors. He has supported the DMSP Cal/Val effort by developing detailed simulations of the SSMIS and DMSP Block 5D-3 spacecraft operation. Gene A. Poe (M 91) received the B.A. and M.S. degrees in electrical engineering from the University of California, Berkeley, in 1964 and 1965, respectively. He has worked in wide-ranging capacities for major aerospace companies responsible for the development of space-based passive microwave instruments (Aerojet Corporation, and ; Hughes Aircraft Company, ). His experiences in microwave radiometry include analyzing laboratory and field measurements, developing emissivity models, and participating in the design of space sensors (DMSP SSM/I, SSMIS, SSM/T- 2, and NOAA AMSU). From 1986 to 1989 he worked at the Space Sensing Branch, Naval Research Laboratory, Washington DC, on passive microwave modeling and analysis of SSM/I data. In 1993 he rejoined the Naval Research Laboratory to lead the Calibration/Validation program for the SSMIS instrument and currently works in the Satellite Meteorology Branch in Monterey, CA.