As a result of the Clean Air Act Amendments of 1970 and 1990, there have been significant decreases in sulfate (SO42-) concentrations in surface waters across the northeastern United States. The 37 Direct/Delayed Response Program (DDRP) watersheds in the Adirondacks receive elevated levels of atmospheric S deposition and showed considerable variability in lake SO42- concentrations. In response to decreases in atmospheric S deposition, these sites have generally exhibited relatively uniform decreases in surface water SO42- concentrations. In this study, an integrated biogeochemical model (PnET-BGC) was used to simulate the response of lake SO42- concentrations at these DDRP sites to recent changes in atmospheric S deposition. Using default parameters and algorithms, the model underpredicted lake SO42- concentrations at sites with high SO42- concentrations and overpredicted at sites with low SO42- concentrations. Initial predictions of lake SO42- were relatively uniform across the region. Initial model simulations also underpredicted decreases in lake SO42- concentrations from 1984 to 2001. We identified seven hypotheses that might explain the discrepancies between model predictions and the measured data. Model inputs, parameters, and algorithms were modified to help test these hypotheses and better understand factors that control spatial and temporal patterns in lake SO42- in this acid-sensitive region.