Modeling soil organic matter (SOM) in forest soils is challenging due to temporal variations (e.g. changes in climate and disturbance events) and spatial variations (e.g. changes in topography, drainage and elevation). Two computer models plus a landscape-based regression model were used to address these challenges in the hardwood forests established on acid-tilled soils in the Green Mountains of Vermont. Data was gathered from 78 hardwood stands in a study conducted by Boyd W. Post and Robert O. Curtis in 1957-1960. Additional data was gathered when the University of Pennsylvania revisited 41 of the 78 stands in 1990-1992. No significant change in the average SOM for the whole region was observed over the 33 year period.

For individual stands of the 1992 dataset, above-ground biomass and stand age did not correlate with SOM, suggesting that SOM in this forest is not strongly coupled with vegetation characteristics. However, in a stepwise multiple regression analysis, soil depth, slope, latitude, elevation and available moisture predicted total solum SOM (r2 = 0.45). In a computer modeling study, the PnET and Century models were tested and compared to a benchmark average (average SOM for the top 20 cm of mineral soil plus the forest floor for all the 1992 stands - 23,300 g/m2). Initial runs showed that both PnET and Century underestimated SOM by about 20-40 percent. This study compares three approaches to estimating changes in SOM for a northern hardwood forest.