Accurate forest mortality monitoring is important for understanding ecosystem dynamics and sustainable forest management. Standing dead-tree volume fraction is a useful indicator of forest condition and biodiversity, but estimating it beyond field plots remains challenging. Remote sensing can support plot-level prediction, yet it is unclear whether satellite embeddings can substitute aerial multispectral imagery (MSI) or add information when combined with canopy height model (CHM) and MSI predictors. Satellite embeddings (SEs) are precomputed learned feature representations from the AlphaEarth Foundations model, which summarizes multi-source satellite observations into annual 10 m high-dimensional embedding layers. SEs are consistently available across broad areas and encode spectral, spatial, and temporal information beyond conventional single-date predictors. Using 134 field plots from primary and near-natural boreal forests in Finland, we predicted standing dead-tree volume fraction using CHM, aerial MSI, and SE predictors. Random Forest, Gradient Boosting, and XGBoost models were evaluated across seven predictor-set combinations using nested cross-validation. Main comparisons included CHM+MSI, CHM+SE, and CHM+MSI+SE. For XGBoost, CHM+SE nearly matched CHM+MSI performance (R²=0.675 vs. 0.679), suggesting SEs may substitute MSI when MSI is unavailable. The highest accuracy was achieved with CHM+MSI+SE (R²=0.703±0.073, RMSE=0.100±0.018, MAE=0.075±0.012), indicating a modest benefit from combining all predictor sources. SHapley Additive exPlanations (SHAP) analysis revealed that SE predictors interacted with canopy-structure and spectral variables, with temporal-change embedding predictors among the most influential. SEs can support plot-level standing dead-tree volume fraction prediction, particularly where aerial MSI is unavailable. Future work should address spatial transferability, longer embedding trajectories, and wall-to-wall forest mortality mapping.