|
Data collected with ground-based telescopes accounts for the overwhelming majority of astrometric observations of mainbelt and near-Earth asteroids. Earth’s atmosphere subjects these measurements to random error from seeing and to systematic bias from differential color refraction (DCR). The DCR bias when observing solar-illuminated targets with nonuniform spectral reflectances and using non-solar-analog stars as fiducials can be several tens of milliarcseconds, even at low airmass. The direction of DCR bias is aligned with local vertical at the observing telescope and its varying orientation in inertial space masks its signature in aggregate error analysis performed in inertial coordinates. Until recently, DCR effects of tens of milliarcseconds were dominated by the hundreds of milliarcseconds of systematic bias present in astrometric star catalogs. Improvements in the accuracy of catalogs beginning in 2015 with the 30 milliarcsecond URAT1 catalog, the 2017 publication of the 25 milliarcsecond UCAC5 catalog, and the forthcoming sub-milliarcsecond GAIA catalog have lowered the error floor on achievable accuracy to the point where DCR is now the dominant systematic bias in data taken from the ground. DCR bias depends on the spectral quantum efficiency of the observing instrument, the spectral reflectance of the target, and the spectral types of the fiducial stars. To realize the benefit of star catalogs accurate below the 30 milliarcsecond level, spectroscopic measurements of asteroids and fiducial stars are necessary to correct for DCR bias. We analyze archival observations of asteroids with known spectral types and present new findings with our own highvolume observations of GPS and GLONASS satellites and the asteroids 4179 Toutatis and 3122 Florence reduced using the URAT1 and UCAC5 catalogs to show that DCR, rather than catalog bias, is now unambiguously the dominant source of systematic error in ground-based astrometry. Our observations of 3122 Florence with the r’ and i’ passbands exhibit vertical residuals more than 100 milliarcseconds beyond what we predict using published reflectance spectra. We attribute the discrepancy between prediction and measurement to the high sensitivity of predicted DCR bias to the slope of the asteroid’s spectral reflectance within the r’ and i’ passbands and caution against relying on narrow passbands alone to mitigate DCR bias. We derive requirements for measurements of instrument spectral quantum efficiency and asteroid spectral reflectance necessary to compensate for DCR to a level commensurate with the accuracy of modern catalogs. The instrument passband must be well-sampled, and while a spectral resolution of 75 nm is sufficient on average when using an unfiltered silicon detector, a resolution of 10 nm is required to ensure worst-case astrometric accuracy of 25 milliarcseconds when observing asteroids with Sloan passbands down to zenith distances of 75 degrees. Systematic biases of tens of milliarcseconds correspond to kilometers of instantaneous cross-range error when observing asteroids. For certain geometries, that error builds with more data rather than averaging out. We examine all likely Earthimpact scenarios and find that when an asteroid approaches from inside 1 AU and forces ground-based observations to occur at large zenith distances, lack of DCR compensation results in impact point predictions that are biased significantly away from their true locations. We present hypothetical NEO discovery scenarios where at fewer than four months before impact, the bias in impact point estimates derived from available ground-based data is many hundreds of kilometers beyond the five-sigma formal uncertainty of the estimate.
|
