Legally protected national parks provide an appropriate substrate for essential long-term study of ecosystem structure and function, and for detecting trends in natural and human-induced stress. The absence of unplanned site manipulation in such areas is especially valuable for such research. Our present research has two major components. The first is the long-term ecosystem-level study of the effects of atmospheric contaminants on ecosystem processes. The overall objective is to evaluate ecosystem aquatic/terrestrial linkages and their role in establishing aquatic ecosystem sensitivity to anthropic atmospheric inputs. Four watershed/lake ecosystems, representative of much of the region's diversity, are under study. Two mature boreal sites on Isle Royale are characterized by first-order perennial surface stream input and lake outflow. Two additional mainland northern hardwood sites, one with shallow soils and one with soils derived from glacial till, are characterized by sensitive aquatic systems. One site is in a private reserve and the other in Pictured Rocks National Lakeshore. Surface outflow is gaged by Parshall flume and stage height recorder. Meteorological stations record variables for estimating evapotranspiration. One-tenth ha plots have been established in all watersheds and three sites have had intensive study of precipitation modification by canopy and forest soil. Five-year mean maximum and minimum lake pH varies from 6.85 to 4.94, Ca(2+) from 1070 to 54 μ eq l(-1), K(+) from 5.42 to 8.35 μ eq l(-1), NH 4 (+) from 10.12 to 3.23 μ eq l(-1), HCO 3 (sup-) from 635 to 24 μ eq l(-1), NO 3 (sup-) from 3.27 to 1.54 μ eq l(-1), and SO 4 (sup2-) from 110 to 52.7 μ eq l(-1). The relatively high NO 3 (sup-) values observed in one lake are the result of stream drainage from a watershed dominated by Alnus rugosa, and another has high seasonal NO 3 (sup-) inputs during spring runoff. However, owing to periodic winter thaws, significant snowpack release of nutrients generally precedes maximum spring stream runoff. Water chemistry in both sensitive and non-sensitive lakes appears to be primarily reflecting how the conterminous terrestrial system is retaining atmospheric inputs more than the quality of direct lake atmospheric input. This is especially evident for H(+), NO 3 (sup-) and SO 4 (sup2-) .The second component is the assessment of watershed acidification, SO 4 (sup2-) output and soil retention across an input gradient. An anthropic deposition gradient provides the opportunity for intersite time-trend analyses as to the effects of inputs. Our study objective was to see if the decreasing west to east input/output values for SO 4 (sup2-) , noted in small first-order watersheds in national parks from Minnesota to Ohio, might be related to present atmospheric inputs, potential and total soil SO 4 (sup2-) adsorption, or soil SO 4 (sup2-) desorption from earlier higher inputs. Precipitation pH ranged from 5.05 at Fernberg, Minnesota to 4.24 at Wooster, Ohio. Minimum and maximum concentrations of NH 4 (+) , NO 3 (sup-) , SO 4 (sup2-) and Cl(-) were also found at these stations. Stream water concentrations of NO 3 (sup-) and SO 4 (sup2-) increase in a similar but sharper gradient. Streams are well buffered. Cation, HCO 3 (sup-) , NO 3 (sup-) and especially SO 4 (sup2-) output increase west to east, but H(+) output decreases. At the eastern site stream SO 4 (sup2-) concentration and output exceed HCO 3 (sup-) . Potential soil SO 4 (sup2-) adsorption capacity increases eastward, but this capacity is filled. Crystalline Fe hydrous oxides appear more effective than amorphous Fe hydrous oxides at adsorbing SO 4 (sup2-) . High anthropic anion inputs, inability of forest soil to adsorb additional inputs and perhaps SO 4 (sup2-) desorption appear responsible for the replacement of HCO 3 (sup-) by SO 4 (sup2-) in stream water. The major cation accompanying SO 4 (sup2-) is Ca(2+).