In the context of water resource management, it is known that the direct measurement process of flow rates in natural watercourses tends to be costly and time-consuming, making it common to work with indirect measurements by the so-called rating curve, which is not error free. Therefore, the general objective of this work was to evaluate the applicability of the Bayesian method called Stage-Period-Discharge (SPD) for assessing rating curves in hydraulically unstable river gauge stations, comparing two different approaches to heteroscedastic modeling of structural errors existing in the literature, as well as evaluating the information transfer capability between validity periods. To this end, a case study was conducted at UHE Peti Carrapato, a station with a high degree of hydraulic instability operated since 1954 and with 14 validity periods. The results indicated that a linear error model was the most suitable because it was more parsimonious. It was also observed that the transfer of information between validity periods, although it showed gains relative to prior knowledge, was impacted by significant errors in the context of prediction, which can be explained by the high degree of instability of the selected station.