Dual-dimension spectral reconstruction: spatiotemporal mapping of soil organic matter under complex cropland surface conditions

Accurate soil organic matter (SOM) monitoring is vital for agricultural sustainability. In Shandong Province, soil moisture heterogeneity and surface cover severely limit spectral inversion. To address this, we propose a ‘temporal moisture reduction - spatial unmixing’ strategy for high-resolution SOM mapping. This approach achieves dual spectral reconstruction through minimum soil moisture (MinSM) synthesis and a vegetation-crop residue reduction model. Using the Random Forest (RF) model, we achieved optimal inversion (mean R2=0.76±0.01, mean RMSE=2.25±0.03g/kg) and generated annual 10-m resolution SOM maps (2015–2025). Results reveal a spatial pattern with enrichment in northwestern plains (>16.50 g/kg) and depletion in southeastern areas (~15.00 g/kg), primarily driven by straw return and parent material. Temporally, SOM followed a trajectory from rapid policy-driven increases to high-level stabilization. This framework effectively overcomes complex surface interference, offering a generalizable technical paradigm for fine-scale soil monitoring in complex agricultural ecosystems.