1. where m is the first order gain, q is the second order gain, is the observed blackbody counts, is the observed space counts, is the radiance of the blackbody, is the radiance of the scan mirror and are the emissivities of the scan mirror at the observation angles of the blackbody and space looks respectively. There are two things to note: emission from the scan mirror is included as a self emission term and for the 3.9 µm detector q =0 i.e. the detector is assumed linear. Since the temperature of the IR detectors are closely controlled to 94 or 101K the 3.9 µm gain should not change significantly over short (days→weeks) timescales. INFRARED SENSOR PHYSICAL BIAS CORRECTION Jonathan Mittaz 1, Andrew Harris 1, Eileen Maturi 2 1 Cooperative Institute for Climate Studies, ESSIC, University of Maryland 2 NOAA/NESDIS, Camp Springs, Maryland We have studied the bias correction problem by looking at the 2005 GOES-12 matchup data. Figure 1 shows the 11 µm bias as a function of the atmospheric correction showing the general square-root curvature which is characteristic of a spectral response function error. However, the deviation should be zero when atmospheric correction is zero, and should increase as atmospheric correction increases. Therefore the presence of both a positive offset and a negative slope indicates a problem with the data. The most likely culprit is calibration. To investigate this further, we have also plotted the bias as a function of time in the day which is shown in Figure 2. Note that trends exist in the bias at all times (both day and night) and at high wind speeds so the effect cannot be dominated by diurnal warming. There are also strong systematic deviations in the bias (the strongest is at 2000 hours which shows the presence of a warm tail in the bias distribution). Neither the overall bias variations nor odd events such as the hot tail seen at 2000 hours have an obvious geophysical explanation. As with Figure 1, the most likely explanation are errors with the calibration. Figure 2. The left hand panel shows the difference between 11µm BTs and the expected BTs from a CRTM simulation using the relevant NCEP model values as input. As in Figure 1, the simulated BTs have been corrected for any difference between the buoy SST and NCEP model SST. Also shown as coloured lines are the biases when the data is filtered with respect to wind speed. There may be a possible diurnal signal but the overall bias trends exist even at high wind speeds where there should be no diurnal warming. The middle panel is the standard deviation for the complete dataset and shows a large peak around 2000 hours. The rightmost panel investigates this more fully plotting the actual distribution of BT differences for the two marked times and shows the presence of a pronounced warm tail in the 2000 data. Most of the bias effects shown in these plots are difficult to explain geophysically and indicate that calibration problems may be important The calibration system on GOES-12 consists of a sequence of space look observations taken every 2.2 seconds and blackbody observations every 30 minutes. The first-order gain is then derived using Figure 3. Top two panels shows the first order gain derived using the standard calibration equations with no correction for the MBCC for the 3.9 and 11µm bands. The bottom panels show the first order gains for the same filters taken directly from the GVAR calibration data which has the MBCC effect removed. There is one extra known calibration effect called Midnight Blackbody Calibration Correction (MBCC * ), that has to be considered. The MBCC is thought to be caused by radiation from the Image Navigation and Registration sunshields being reflected by the blackbody during the calibration observations. This effect is meant to be removed in the current GOES processing (see Figure 3). Figure 4. The leftmost panel shows the variation in both the blackbody and scan mirror temperatures and shows that the scan mirror temperature leads the blackbody temperature. This gives rise to a characteristic loop in the scan temperature/blackbody temperature plane shown in the second panel. The rightmost panels shows the GVAR gains for the 3.9 and 11µm filters plotted as a function of blackbody temperature. A loop can be seen in both filters, indicating that there Is some residual emission not removed by the MBCC algorithm. In order to investigate this further, we have simulated the calibration system allowing for the addition of extra emission terms which we have related to the temperature of the scan mirror. The simulations are plotted in Figure 5 and show that adding extra emission to the blackbody counts can indeed give you loops in the gain-temperature plane. In the case of no extraneous emission the 11µm gain shows a monotonically decreasing value as a function of temperature while the 3.9 µm gain is constant. This is the expected behaviour if the calibration is correct. Figure 6 shows the case for the 3.9 (linear) detector. Adding in extra emission of order 0.055 can remove the loop from the gain/temperature plot but it still leaves a large variation in the 3.9 µm gain which really should be constant. Therefore using the scan mirror as a proxy for the extra emission does not seem to work very well, though given that the true source of emission is thought to be the sunshields this may not be too surprising. Figure 5. Simulations of the effect of extra emission on the first order gain. When extra emission is present characteristic loops in the gain/BB temperature plane are seen. With zero extra emission the 3,9µm gain is constant whereas there is a slope in the 11µm gain caused by the non- linearity of the 11µm detector Is all the MBCC removed? The MBCC is known to be strongly correlated with the temperature of the optical system. Since this temperature leads the blackbody temperature there will be an observable effect in the gain vs. blackbody temperature plane if the MBCC is not removed completely. The two right hand figures in Figure 4 show the current operational GOES gains plotted against BB temperature and show a loop caused by missed MBCC emission. Therefore as a first attempt at fixing the problem with the GVAR calibration we have looked at simply adding in an extra emission term phased with the scan mirror to the calibration equation The initial motivation for this study comes from the application of physically-based retrieval methodologies to the GOES-Imager to generate sea surface temperature (SST). Empirical algorithms derived by direct regression can eliminate biases in a mean sense but only for retrieval conditions encompassed within the matchup dataset used. Physically-based retrieval methodologies, on the other hand, allow for better specification of errors in remote regions devoid of in situ data by using ancillary data (e.g. from NWP analyses) and fast-forward radiative transfer models, this also increases the information available which leads to better SST retrieval accuracy. Such radiative transfer based methodologies are reliant on accurate sensor calibration and characterization, i.e. it is important that discrepancies between modeling and observation are only stochastic in nature. INTRODUCTION GOES-12 CALIBRATION AND THE MBCCA MODEL OF THE CALIBRATION SYSTEM CONCLUSION We have shown that the radiance bias correction for GOES-12 is still an issue as shown by the clear residual problems in both the GOES-12 data and its calibration. A final solution is not yet available but there are indications that modifying the calibration equations to include extra emissive terms may solve the remaining problems. Thus, while attempting to remove the MBCC and associated emission by modifying the calibration equations directly would seem to have merit (and indeed may be necessary), more work Is needed to determine how to parameterize different self-emission components involved in order to find a solution to the GOES-12 calibration problem. Figure 1. Plot of modeled – observed brightness temperature for the 11 µm channel of GOES-12 Imager versus the modeled atmospheric correction (i.e. T sfc – BT 11 ). The modeled BT has been adjusted to account for sub-NCEP-grid SST variability by using the in situ (buoy) SST and the tangent linear operator supplied by the CRTM. The right hand plot shows the same bias as a function of 11µm BT alone and the large feature centered around 295 K actually covers the full range of variability in radiance bias showing that there is little skill in resolving this region by using observed T11 as a predictor * Johnson, R, and M. Weinreb, 1996: GOES-8 Imager Midnight Effects and Slope Correction. In GOES-8 and Beyond, Edward R. Washwell, Editor, Proc. Society of Photo-Optical Instrumentation Engineers (SPIE), 2812, pp. 596-607. Figure 6. Left hand panel shows the 3.9µm first order gain as a function of Blackbody temperature showing a clear MBCC loop. The right hand panel shows the same data but with an extra emission term add to the calibration which has removed the loop but has still left a large variation in gain inconsistent with the linear nature of the 3.9 µm detector. GOES-12 RADIANCE BIASES