Mapping soil organic carbon (SOC) in salt marshes is essential, as not only it indicates soil health, but also is a proxy for marsh resiliency against rising sea level and degrading belowground biomass. However, SOC mapping efforts have been hindered by the low availability of SOC data due to difficulty in field sampling and longer processing time. We collected around 250 SOC point measurements and developed a Landsat -based machine learning (ML) model to map SOC from 2000 – 2024. We emphasized learning long-term environmental processes via ML models that affect SOC dynamics in salt marshes. This was achieved by carefully choosing model input data that had similar patterns of covariation with SOC for data from 2022 – 2024 (majority of data: 250 points) and data from 2000 – 2024 (70 additional points from legacy datasets). Preliminary results indicate that SOC can be predicted with around 65% accuracy and 10% error. Moving forward, we will explore more inputs for the ML model and will evaluate model accuracy fluctuation over the study period. Landsat-based multidecadal SOC estimates would be very useful for the larger scientific community working in the region and would improve understanding of SOC changes and its drivers.