Abstract
Mismodeling of solar radiation pressure (SRP) is a leading error source in BeiDou precise orbit determination (POD), yet the temporal predictability of the empirical SRP parameters remains poor. Using the five-year (2020–2025) daily ECOMC parameter series of 42 BeiDou-2/BeiDou-3 (BDS-2/BDS-3) satellites (67,824 records) and driving attribution, we build orbit-group- and manufacturer-specific prior-constraint and parameter-reduction strategies. The BDS-2 GEO YSIN is almost linear in the solar elevation angle β; the BDS-3 IGSO YRAD follows a second-order harmonic in β; and the BDS-3 MEO-CAST (China Academy of Space Technology) DCOS is well described by a joint cosβ and eclipse-season-offset prior (\({R}\)2 = 0.814). These priors are purely empirical statistical models, and the IGSO group contains only three satellites, so transferring these priors to newly launched satellites of the same type carries risk. Full-year 2023 five-scheme POD validation shows the priors are neutral relative to free estimation, whereas indiscriminate parameter reduction degrades the within-day fit by a factor of 7.35; the recommended partial reduction for the BDS-3 IGSO and BDS-3 MEO-SECM (Shanghai Engineering Center for Microsatellites) satellites avoids this degradation. For the CAST group, reduction is not recommended: a control experiment shows the DCOS prior cannot recover the accuracy lost to reduction. Parameter-domain statistics alone cannot guide reduction; orbit-domain POD validation is mandatory.