Annual catchment export of total phosphorus (TP), dissolved organic carbon (DOC), total iron (Fe), and dissolved organic nitrogen (DON) to seven lakes in central Ontario was measured between 1978 and 1998. Fluctuations in annual water discharge and total DOC load (including precipitation) to the lakes over the 20-year period were similar in the seven study lakes. DOC export to the lakes responded proportionally to changes in discharge, decreasing during drier and warmer years. There were similar but less accentuated variations in annual DOC lake concentrations. There were no clear regional trends evident during the 20-year period toward drier or wetter conditions, less DOC load, clearer lakes, etc., that could be interpreted as signaling a shift toward a different equilibrium state. The fraction of the DOC load retained by lakes (transferred to sediments and the atmosphere) increased during an extended dry period. Fe, TP, and DON export decreased more than DOC export during the extended dry period. Runoff appears to affect Fe, TP, and DON export first by controlling export of organic matter and second by affecting water table position and thus redox levels in the surface layer of peatlands. Permanently drier conditions with less runoff would likely lead to clearer lakes that are less productive. Conversely, increased runoff would lead to more colored and productive lakes. Both scenarios have implications for subsistence and sport fishing economies.
 Several multiyear studies have reported changes in lake and riverine DOC concentrations over time. DOC concentrations increased significantly between 1988 and 2000 at 20 of 22 acid and non-acid sites in the UK Acid Waters Monitoring Network [Freeman et al., 2001]. Significant increases in color but no change in DOC concentrations were observed in fall samples taken from 24 lakes in southern Norway between 1996 and 2001 but both DOC and color increased in a water treatment plant intake between 1989 and 2002 [Hongve et al., 2004]. Increases in DOC and color have also been reported for surface waters in southeastern Norway, southern Sweden, southwestern Finland, central Europe, and the northeastern United States (data not shown but cited by Nordtest ). The absorbance data of Andersson et al.  suggest DOC oscillations in Swedish rivers with a 10-year cycle between 1965 and 1985 whereas color increased linearly or exponentially between 1967 and 1988 in four Swedish lakes [Forsberg and Petersen, 1990]. In contrast, DOC was relatively constant in three lakes in northwestern Ontario between 1970 and 1980 but declined significantly between 1981 and 1990 [Schindler et al., 1997]. All that can be said with certainty is that trends, if any, depend on the length of the study, the specific years, and the geographic location. Changes in analytical methods must also be taken into account.
 Fluctuations in DOC concentrations and catchment export have typically been attributed to changes in runoff. However, Freeman et al.  argued that changes in runoff do not account for the UK increases (analyses were not shown) and that increases in air temperature could have. This has led to some discussion of the nature and complexity of regional-scale climatic factors [Tranvik and Jansson, 2002; Evans et al., 2002]. In an experimental study that manipulated temperature and water level in a bog and a fen, Pastor et al.  concluded that discharge rate was the most important variable. DOC concentrations in the bog increased slightly with heating but were unaffected by heating in the fen.
 This paper examines temporal variation in catchment export of DOC and several related substances, total phosphorus (TP), dissolved organic nitrogen (DON), and total iron (Fe), to seven study lakes in the Dorset region of central Ontario between 1978 and 1998. Several questions are asked. Are there long-term trends in DOC export to lakes and the climatic factors: runoff and air temperature? Do runoff and air temperature appear to control DOC export? Does catchment export of Fe, TP, and DON respond similarly to DOC export or are different factors involved? This paper also examines long-term trends in lake retention of DOC.
2. Study Sites
 The Dorset study lakes (Table 1) have been the focus of a long-term study of the impacts of long-range atmospheric transport, climate change, and cottage development on water quality in forested headwater catchments and lakes. The seven study lakes and their twenty subcatchments are located in the district of Haliburton-Muskoka on the southern edge of the Precambrian Shield in central Ontario (Figure 1). The streams are first and second order. The lakes are oligotrophic to mesotrophic headwater lakes with the exception of Red Chalk Lake which receives discharge from Blue Chalk Lake.
Table 1. Characteristics of Study Lakes Near Dorset, Ontario, Between June 1978 and May 1998a
Lake area, Ao (ha); catchment area exclusive of lake, Ad (ha); mean depth, zm (m); and mean, whole lake TP concentration (μg L−1); pH and DOC concentration (mg L−1).
TP concentrations not available for Red Chalk in 1981/1982 and 1982/1983.
 The catchments are primarily forested with some cottage development and are underlain by Precambrian metamorphic plutonic and volcanic silicate bedrock. The depth of overburden to bedrock is less than 10 m in almost all locations, and generally less than 1 m. The extent of peatlands ranges from 0 to 25% of the stream catchment areas. The lakes and catchments are described in more detail by Dillon et al. .
 Stream, lake, and precipitation samples were analyzed for DOC. DOC was measured by acidifying samples and flushing with nitrogen gas to remove inorganic C. Organic carbon was then oxidized to CO2 by exposure to UV light in acid-persulfate media and colorimetrically measured with phenolphthalein, calibrated with phthalate standards. Total phosphorus (TP) was measured using ascorbic acid-molybdate colorimetry after autoclave digestion in sulphuric acid. Stream samples were analyzed for total Fe colorimetrically, after autoclave digestion in acid (30 g of hydroxylamine hydrochloride with 200 mL sulphuric acid in 1 L of distilled water), for formation of the ferrous-2,4,6-tris(2'pyridyl-)1,3,5-triazine complex. Precipitation samples were analyzed for total iron with inductively coupled plasma using argon and atomic absorption spectrophotometry. Analytical methods for DOC, TP, and Fe are described by Ontario Ministry of the Environment .
 Samples were collected from stream and lake stations at 1- to 4-week intervals, filtered through 80-μm polyester mesh into pre-rinsed Nalgene bottles (DOC) or 50 mL pyrex culture tube (DIC), and placed in temperature-controlled containers while in transit to the laboratories [Dillon et al., 1991; Molot and Dillon, 1993]. Water level or stage was recorded continuously at weirs or flumes installed on the study streams [Scheider et al., 1983]. Stage discharge relationships were constructed for each stream.
 Annual export from gauged streams was calculated by multiplying the concentrations in the streams (measured during the midpoint of a time period) by the total discharge over the time period and summing for the hydrological year (1 June to 30 May). DOC export per square meter from ungauged portions of the catchment was assumed to be identical to export from the gauged portions. Maximum daily air temperature was monitored at between one and four meteorological stations located at three of the Dorset study lakes and the Dorset laboratory.
 Bulk precipitation was collected and precipitation depth monitored at up to 12 stations, with never less than four stations in operation. Precipitation samples were removed from collectors (0.25 m2) with Teflon-coated, stainless steel funnels. Collection periods ranged from 1 to 40 days, although samples were typically removed weekly.
 Lake samples were collected as integrated, volume-weighted epilimnetic, metalimnetic, and hypolimnetic samples, with the appropriate depths determined from temperature profiles taken during summer stratification, or in tube composites at all other times of the year. Whole-lake concentrations for any sampling day during stratification were calculated by mathematically weighting the concentrations in the epilimnion, metalimnion, and hypolimnion according to the relative proportions of each strata.
 Results are presented here for the period June 1978 to May 1998 with the following exceptions. DOC concentrations were not measured in Red Chalk in 1980/1981 and 1981/1982; hence mass transfer coefficients cannot be estimated for these 2 years (see model below). Sampling of Dickie Lake was halted in May 1995. Stream sampling began in Crosson in June 1980 and was discontinued in April 1993. Sampling of Plastic Lake began in June 1979.
 A steady state model was applied that predicts mean annual DOC concentration in lakes as a function of inputs, outputs, and flushing rate [Molot and Dillon, 2003]. It is a one-box, continuously stirred model that assumes the lake is extremely well mixed, material influx and efflux are constant over time, and all DOC sources to the lake are external. A steady state equation relating inputs and outputs for DOC is given by
where l is load, O is outflow, L is loss, S is sediment storage, and M is mineralizationAssuming that outflow and loss are reasonably described by first-order kinetics,
At steady state,
where [DOC] is the mean annual DOC concentration in lake water, l is the total annual DOC input from the catchment and precipitation, O is the annual loss via the lake outlet divided by lake area (mass m−2 yr−1), qs is the areal water discharge rate from the lake (m yr−1), vdoc is a net generalized loss coefficient (m yr−1) which integrates loss of DOC via burial, photo-oxidation, and mineralization, and R is the fractional retention. Export refers to the total load to the lake minus the contribution from precipitation. Data are presented primarily as 3-year running means normalized with the long-term mean to reduce annual noise and facilitate between-lake comparisons.
 Fluctuations in 3-year means for annual discharge over the 20-year period (Figure 2) were similar in the seven study lakes in response to regional-scale climate variation. A drying trend that began in 1983/1984–1985/1986 lasted until 1988/1989–1990/1991 and was followed by increases until the last 3-year period in 1995/1996–1997/1998. The 3-year normalized running means for discharge during the driest period in 1986/1987–1988/1989 and 1987/1988–1989/1990 were 0.74–0.87 of the long-term mean, compared to 1.14 to 1.24 during the wettest period in 1994/1995–1996/1997, a difference of about 30%. The dry periods were accompanied by warmer maximum daily air temperatures in June–October.
 Long-term mean annual l, O, R, qs, and vdoc are presented in Table 2. Mean annual DOC load ranged 5.2-fold from 8.0 g m−2 yr−1 in Blue Chalk to 41.8 g m−2 yr−1 in Chub Lake reflecting differences in the amount of peatland and ratios of lake surface area to catchment area [Dillon and Molot, 1997b]. Mean annual retained DOC ranged 3.9-fold from 5.0 g m−2 yr−1 in Blue Chalk to 19.6 g m−2 yr−1 in Dickie Lake (Table 2).
Table 2. Mean and Standard Deviations of Long-Term Areal Water Load, qs, (m yr−1), DOC Load, l, and Outputs, O (g m−2 yr−1), per Lake Areaa
R, fraction of the load retained (not discharged); vdoc, mass transfer or loss coefficient (m yr−1). The first line is the mean with the standard deviation below; r is the Pearson correlation coefficient for within-lake annual areal water discharge, qs, and annual DOC load.
 The mass transfer coefficient, vdoc, is of interest because it indicates the efficiency with which internal mechanisms transfer DOC to lake sediments as POC and to the atmosphere via respiration and photo-oxidation (mostly as CO2) without being obscured by differences in residence time and load. Long-term means were similar among the lakes, ranging 1.5-fold among the non-acidified lakes from 2.8 m yr−1 in Blue Chalk to 4.2 m yr−1 in Red Chalk and 5.2 m yr−1 in atmospherically acidified Plastic (Table 2). The vdoc varied 2.0- to 2.9-fold between years within-lake with the lowest value of 1.7 m yr−1 in 1981/1982 in Blue Chalk and the highest 6.6 m yr−1 in Red Chalk in 1992/1993 among the non-acidified lakes. Until annual variation in vdoc can be predicted, this parameter should be calibrated with a multiyear data set, preferably with more than one lake, and should only be used to predict mean, long-term fluxes.
 The responses of total DOC load (which is primarily export from catchments because the precipitation contribution is very small) and retained DOC (DOC load not discharged downstream) were similar to discharge (Figures 3a and 3b) although somewhat noisier. In contrast to DOC trends in the United Kingdom [Freeman et al., 2001], DOC export, and volume-weighted stream concentrations decreased during drier, warmer years, for example, Harp tributaries (Figure 3c).
R, the DOC fractional retention (equation (3)), (3-year running means normalized to the long-term mean) increased during the dry period but remained high when the dry period ended (Figure 4). The increase in R during the dry period may be explained in part by longer residence times: smaller qs in the denominator of equation (4) increases R at constant vdoc. Since vdoc was not constant and decreased during the dry period in all lakes except Crosson (data not shown), the mathematics of equation (4) indicates that changes in qs are relatively more important to changes in R than changes in vdoc. It is not possible to tell whether partitioning of retained C between sediments and the atmosphere varied during the study period.
 Variations in annual lake DOC concentrations ([DOC]) were similar to load but less accentuated with the 3-year running means dropping to 0.90 of the long-term mean (Figure 5). The dampened [DOC] fluctuations can be explained by examining equation (2b). Using long-term mean values for Harp Lake to illustrate, equation (2b) predicts that lake [DOC] will decline 16% if there is a 30% decline in DOC load from 30.9 g m−2 yr−1 and a corresponding decline of 30% in qs from 4.2 m yr−1 at a constant vdoc of 3.4 m yr−1.
 There were no clear regional trends evident during the 20-year period, for example, no trend toward drier, warmer, or wetter conditions, less DOC load, clearer lakes, etc., that could be interpreted as signaling a shift toward a different equilibrium climatic state. Since DOC load is primarily a function of runoff [Dillon and Molot, 1997a] and extent of peatland in the Dorset area [Dillon and Molot, 1997b], similarities between qs and load suggest that the extent of peatland coverage has not changed significantly during the 20-year study period.
 The correlation coefficients between DOC load and qs for each lake were high (Table 2). The slope of relative annual DOC catchment export versus relative annual discharge was 0.86 (Figure 6) indicating that load is primarily a linear function of runoff [Dillon and Molot, 1997a]. Slopes for each lake ranged from 0.72 to 1.05, none of which were significantly different from 1.0 at the 1% level. If TP load is similarly affected by the same factors that affect DOC export, then the ratio of TP load to DOC load should not deviate significantly from 1 over time. However, a significant drop in the TP/DOC ratio in export (total load - direct deposition) occurred during the extended dry period (Figure 7a), similar to the variation in the ratio of lake concentrations (Figure 7b).
 Peatlands are a major source of DOC and TP as well as Fe [Dillon and Molot, 1997b]. As occurred with catchment export of TP, catchment export of Fe dropped relative to catchment export of DOC during the dry period (Figure 7c). The decline in both Fe and TP export relative to DOC may have been caused by more oxidizing conditions in the peatland surface layer brought on by a lowering of the water table during dry conditions [de Mars and Wassen, 1999].
 The extended dry period also has implications for catchment export of dissolved organic nitrogen (DON). If runoff is the only factor controlling DON export, then DON export will vary 1:1 with DOC. However, the ferrous wheel hypothesis [Davidson et al., 2003] predicts that drier, more aerated conditions will reduce the amount of Fe+2 available for abiotic nitrate reduction to nitrite, leading to less DON formation. In turn, this will reduce DON export. DON export declined during the extended dry period up to 25% below the long-term means while nitrate deposition remained relatively constant (Figure 8a). The decline in DON export was slightly steeper than the decline in DOC export itself in most lakes during the dry period with the exception of Crosson and Chub (Figure 8b). The large decline in absolute catchment export of DON with a small decline relative to DOC export is consistent with the notion that runoff exerts major control on catchment export, first by controlling the leaching rate of organic matter and second by affecting the redox level in the surface layer of peatlands.
 There were no clear regional trends evident toward drier or wetter conditions, less DOC load, clearer lakes, etc., that could be interpreted as signaling a shift toward a different equilibrium state. However, the 20-year monitoring period allows us to cautiously extrapolate system responses during extended wet and dry periods to predict how northern catchments will respond under different climatic scenarios.
 While the available pool of exportable organic matter is generated by numerous biotic and abiotic processes that affect decomposition and leaching, the results from this study suggest that potential increases in decomposition rate brought about by elevated temperatures can be masked by a decline in runoff in this part of the world. The effect of runoff declines on catchment export might be less dramatic in the more maritime climate of the United Kingdom where peatlands might remain inundated even during drier years. We expect that the roughly linear relationship between load and runoff will continue indefinitely in Ontario provided the amount of peatland surface area remains constant.
 Second, the much greater decrease in Fe, TP, and DON loading relative to DOC loading during the extended dry period suggests that mobilization of Fe, TP, and DON from peatlands is sensitive to changes in water table position and redox levels [de Mars and Wassen, 1999]). Drier conditions will therefore result in greater water clarity and reduced productivity downstream, including fish productivity, assuming constant TP and N deposition. However, the decrease in lake DOC and TP concentrations will be less than the decrease in mass export because of the dampening effect predicted by equation (2b). Moreover, we should not expect the extent of peatland, a common feature across the boreal landscape, to remain constant in a permanently drier climate. Conversely, increased runoff would lead to lakes that are more colored and more productive. Both climate scenarios have important economic and social implications for human communities relying on subsistence and sport fishing.
 The last and more speculative point deals with the effect of a drier climate on sediment organic carbon (SOC) accumulation rates, a potentially important issue for countries attempting to meet Kyoto reduction targets through natural carbon sink offsets. The literature is replete with analyses showing that community respiration typically exceeds primary production in unproductive lakes [e.g., Prairie et al., 2002; Hanson et al., 2003], yet lakes accumulate significant amounts of organic carbon. For example, compare SOC accumulation rates of 3.9 to 10.1 g C m−2 yr−1 in the seven Dorset lakes [Dillon and Molot, 1997b] to primary production rates ranging from 18.1 to 51.5 g C m−2 yr−1 (ice-free season only) in Lake 239 in northwestern Ontario [Shearer et al., 1987]. This suggests that there may be an important mechanism other than a “rain of planktonic debris” that transfers organic matter to sediments. One possibility is adsorption of DOC to settling amorphous Fe [Jonsson, 1997], especially recalcitrant humic matter. A drier climate might reduce the importance of this transfer mechanism by reducing Fe loading rates to lakes even more than DOC loading. Therefore, while a drier climate reduces the amount of retained DOC (storage in sediments plus transfer to the atmosphere) [Algesten et al., 2003], it might increase DOC residence time in the water column and divert carbon from storage in the sediments to the atmosphere.
 We thank Joe Findeis for database management.