It has been shown that for asteroid (99942) Apophis the leading source of uncertainty for predictions of its orbital motion is due to non-gravitational accelerations arising from anisotropic thermal re-emission of absorbed solar radiation, known as the Yarkovsky effect . Yet, previous attempts to obtain this parameter from astrometry for Apophis have only yielded marginal detections . Here, we present an independent estimation for the Yarkovsky effect on Apophis from optical and radar astrometry. Our approach is based on automatic differentiation techniques in terms of high-order Taylor series expansions both with respect to time and deviations with respect to a given orbital solution . We implement our own planetary ephemeris integrator, which essentially reproduces the DE430 ephemeris integration . Our dynamical model for Apophis takes into account post-Newtonian accelerations from the Sun, the eight planets, the Moon and Pluto, oblateness effects from Earth's J2 zonal harmonic, perturbations from the 16 most-massive main-belt asteroids and a non-gravitational acceleration term accounting for the transverse Yarkovsky effect. Exploiting these techniques, we implement a Newton method for orbit determination, and perform two orbital fits to optical and radar astrometry: a 6 degrees-of-freedom (DOF) gravity-only orbital fit for the initial conditions, and a 7 DOF orbital fit, which includes the Yarkovsky coefficient A2 as an additional fit parameter, obtaining A2 = (-5.40 ± 3.29)×10-14 au/d-2. Our optical astrometry error model accounts for biases present in star catalogs , and accounts for other sources of systematic errors via an appropriate weighting scheme . Using our orbital solutions, we provide predictions (nominal and 3-sigma uncertainty ellipses) for optical and radar observations for the upcoming close approach that Apophis will have with Earth during 2020–2021 (see Figs. 1 and 2). Finally, we project the orbital uncertainty onto the 2029 B-plane, and propose a parameterized orbit determination scheme, which allows us to compute the time-evolution of the line of variations for Apophis via a high-order Taylor series parametrization.
References:  D. Vokrouhlický et al. (2015). Icarus, 252, 277-283  M. Brozović et al. (2018). Icarus, 300, 115-128  J.A. Pérez-Hernández and L. Benet. (2020). TaylorIntegration.jl v0.8.3. URL: https://github.com/PerezHz/TaylorIntegration.jl  W.M. Folkner et al. (2014). Interplanetary Network Progress Report, 196(1)  S. Eggl et al. (2020). Icarus, 339, 113596  D. Farnocchia et al. (2013). Icarus, 224(1), 192-200.