The objective of this study was to evaluate, modify if necessary, and calibrate a mass balance steady-state total phosphorus (TP) model for use in the Precambrian Shield area of Ontario. Hydrologic data, TP mass balances and lake concentrations data were collected between June 1978 and May 1992 from seven oligotrophic and oligomesotrophic lakes located on the Precambrian Shield in central Ontario. Two of the lakes (Dickie and Harp) had substantial shoreline development; four lakes had alkalinities <25 mu eql(-1). The gross settling or loss coefficient, v, was underestimated when internal load was ignored. Long-term mean rr ranged about three-fold, from 4.7 m yr(-1) in Plastic Lake to 13.7 m yr(-1) in Dickie Lake with an average for six lakes of 7.9 m yr(-1). The exception was Harp Lake where v was 23.2 m yr(-1). The difference in v between heavily developed Dickie and Harp Lakes may have been due to the unusually thick glacial tills in the Harp catchment and their likely ability to retain at least some of the TP from septic effluent. The current inability to predict septic TP retention rates by soils/tills, the highly variable nature of soils/tills, and the decades required to reach equilibrium after installation leave environmental managers with little choice but to apply the ''precautionary principle'' to ensure adequate protection of inland recreational waters and assume that soils do not retain septic TP. There was no evidence that enhanced sedimentation of TP (''oligotrophication'') occurred in any of the three most acidic lakes. In fact, losses via sedimentation were lowest in these lakes. Average retention, R ((input-output)/input) of TP during the ice-free months of May to November was much higher than average R in winter (December-April) in all lakes. Lack of wind-induced mixing, low biological activity and thermal stratification caused by ice cover make it more likely that stream loads are discharged directly from lakes during winter. Seasonal variation in retention suggests that R and v should be higher in warmer climatic zones. There may be no globally valid v, although the data presented here suggest that a regionally applicable estimate of v may be appropriate. The model should only be calibrated with a long-term data set and should be only used to predict long-term, average responses to changes in average TP loading. Copyright (C) 1996 Elsevier Science Ltd
