L band vegetation optical depth (L-VOD) is a widely used remote sensing variable for investigating the spatiotemporal variation in aboveground biomass (AGB). A key step of this method is to fit L-VOD against AGB, then use a space-for-time assumption to infer AGB change from fitted L-LOD change. In this study, we evaluated the performance of different fitting equations and explored their implications in predicting AGB. We used the SMOS-ICV2 L-VOD dataset and four AGB reference datasets. Specifically, we examined the implications of the space-for-time assumption in predicting the AGB interannual variations. We find that all the statistical fitting methods can capture the AGB spatial variation, yet introducing tree cover as a predictor significantly improves the AGB prediction, especially in regions with small and medium L-VOD values. However, these methods all fail to capture AGB spatial variation in dense forests. The AGB reference data also show large discrepancies in these regions. Our results also show that the spatial AGB sensitivities to L-VOD are much larger than the temporal AGB sensitivities to L-VOD, implying considerable uncertainties in temporal AGB changes predicted with spatially built models. By providing a comprehensive evaluation of fitting methods, our results offer a cautionary tale to the use of L-VOD data to infer AGB dynamics and the necessity of developing long-term field-based biomass change datasets for further constraining and evaluating AGB predictions from remote sensing observations.