Data collection

The first step consisted in the establishment of a classification system of the shores based on several criteria that included form of the shoreline (bay, ledge), substratum, topography, and predominant geomorphological processes (erosion, sedimentation). This resulted in the distinction of five geomorphological shore types: depositional flats, floodplains, beaches, terraces, and rock outcrops. Depositional flats are large, flat areas created by the hydrodynamic redistribution of fine mineral sediments over the preexisting surface. High groundwater levels favor the accumulation of thick peat layers. In some cases, wave activity heaped up beach ridges parallel to the shoreline. Floodplains are large, flat areas inundated during periods of high water but without considerable sedimentation in comparison with the depositional flats. Beaches are narrow edges of fine sediment in embayments with moderate slopes. Strong erosion can cause a retrogression of the shoreline creating pocket beaches flanked by rocky ledges. Terraces are raised flats delimited on the landside by a talus of erosion and on the lakeside by a boulder-bordered slope originating from wave breaking. Rock outcrops are mostly steep shores with less than 25 em sediment over bedrock.

The entire perimeter of Lake Duparquet (without islands) was mapped with the object of associating one of these shore types to each homogeneous section of the shoreline with a minimal length of 50 m. Next, 1 06 shoreline points were chosen as locations for the establishment of transects in the unaltered parts of the lake perimeter (Fig. 1.1 ). One hundred points were selected randomly, the number of points per shore type corresponding approximately with the relative length of its total shoreline section2 • To cover all of the rare tamarack stands, six additional transects were drawn through them. The starting point of all transects was positioned relative to a reference water level of 266 m above sea level, which roughly represents the mean level of Lake Duparquet. From each shoreline point, a transect perpendicular to the topographic contours was drawn to analyze the composition of the tree layer of the riparian forest within a 3-m wide strip. An exception was made for the sparse tamarack forests where a 10-m width was used in order to reach a sufficient number of trees. Transects ended at 2 m above mean lake level. The geomorphological situation indicated that it was about the position of high water levels. However, some transects could not be completed up to this height, because the topographic divide between the Jake and adjacent water bodies did not exceed 2 m in elevation. In this case, data collection ended approximately at the culminating point The length of the transects varied between 2 m and almost 400 m . The elevation gradient of each transect was divided into four equal parts, all of these plots covering thus a vertical interval of 50 em. Very long plots were further subdivided until their length did not exceed 20 m in order to get a better resolution of the horizontal distance to the lakeshore. Within each plot, the stem diameter at breast height (dbh) of all living trees larger than 5 em was measured.

Environmental variables

The environmental data comprise the following factors, recorded for all shoreline points: geomorphological shore type, inclination and aspect of the slope, fetch to the opposite shore of the lake, exposure to wave activity, and year of the last forest fire. Type of substratum, elevation, and horizontal distance to the lakeshore, however, were determined for each plot. Inclination and aspect of the slope The inclination of the slope is defined as the angle, measured with a clinometer, between the horizontal at mean water level and the highest point of the transect. The aspect of the slope is represented by the closest oft he eight main cardinal points (N, NE, … , W, NW). Fetch and exposure The energy of the waves arriving at the shore is related to wave height, which is in turn a function of speed and duration of winds, fetch, and water depth (Pond and Pickard 1978). In this study, a modification of the method of Keddy (1982, 1984) was applied using fetch measurements, wind data, and shore slope to calculate exposure measurements. Direct fetch: A map of Lake Duparquet at a scale of 1 : 20 000 was fixed on a drawing board to measure, for each of the 1 06 shoreline points, the distance to the opposite shore of the lake (mainland or island) in 36 different directions (i.e., at 10° intervals). Directions from which waves cannot arrive have a zero value. The mean of this so-called direct fetch for each shoreline point was used as a separate environmental variable (Table 1.1 ). Effective fetch: Narrow embayments and long ledges, but also the many small islands of Lake Duparquet, can result in extreme and less meaningful values for direct fetch (Keddy 1982). Therefore, a correction was performed by taking into account the adjacent fetch measurements: each effective fetch value is composed of the weighted direct fetch values of all adjacent directions that are less than 45o away (for calculations see Keddy 1982).

Wind exceedance: Presuming the shoreline trees are mainly disturbed by severe events, particular importance must be attached to strong winds. Thus, the exceedance value, calculated separately for each of the 36 compass bearings, represents the proportion of strong wind measurements (~20 km/h) in relation to all wind measurements of the same direction based on hourly readings for the time period ranging from 1971 to 1997. The wind data that was used came from the weather station of Rouyn-Noranda, about 50 km southeast of Lake Duparquet. Exceedance was calculated for three different time periods: May, the growing season, and the ice-free period. During May, the ice of Lake Duparquet breaks up and ice scouring increases the effect of wave activity; during the growing season, from June to September, the trees might be particularly vulnerable to wave damage; and during the ice-free period, from May to November, the waves can affect shoreline vegetation. Slope: Waves approaching the shore lose a part of their energy depending on the inclination of the lakebed. To consider this topographical effect, the exposure value was multiplied by the sine oft he slope, assuming that the slope above the waterline is the same as that in the near offshore zone. Thus, the exposure value for a flat shore where the waves are broken by the gently rising lake bottom decreases considerably (e.g., sin(1°) = 0.017 as the multiplier) whereas it doesn’t change much for a steep slope (e.g., sin(60°) = 0.866). By using fetch (direct, effective) and wind exceedance (May, growing season, ice-free period) in combination with or without the correction by the inclination of the lakeshore, 12 (2 x 3 x 2) different exposure values were received for each shoreline point. Six values were calculated by multiplying fetch with wind exceedance and the consecutive summation over all 36 directions:

Data analyses

Only those species present in more than three transects were included in the analyses. Thus, Pinus banksiana, Pinus strobus, and Pinus resinosa were not considered. The major habitat of these three pine species seems to be on the islands of lake Duparquet rather than on the mainland (Bergeron and Brisson 1990, Bergeron et al. 1997). Since the four encountered species of Salix (S. discolor, S. Iucida, S. petiolaris, and S. rigida) were not present in more than three transects and because all of them can be associated with humid habitats (Marie-Victorin 1995), their analysis was limited to the genus level. Of the 1 06 transects, 11 had to be omitted mainly because no living individual of the selected tree species was recorded within the 2-m vertical interval or, if some were present, they were too small to be considered (dbh < 5cm). Plots without any living trees, quite often encountered at the lowest vertical interval (0-50 em) where bare sediment or rock, herbs and shrubs predominate, were not included in the analyses. Hence, the overall data comprise 241 plots from 95 transects (Table 1.1 ). Twenty environmental variables were used as predictors to try to explain the distribution of the 1 0 common tree species occurring in the riparian zone of lake Duparquet. In statistical calculations, the three nominal environmental variables, namely shore type, substratum, and aspect, were transformed into dummy variables. Stepwise logistic regression (SPSS Inc. 1997) was applied to help identify the combinations of factors that explain the occurrence of each species on the lakeshore.

For this purpose, each plot was described in terms of presence or absence of every tree species. Environmental variables were entered in the model if their scores were significant at a level of 0.05, corrected by the method of Bonferroni (Sakal and Rohlf 1995). To deal with the problem of missing values for the fire variable, a first run was performed. If fire was not retained by the forward selection procedure, the calculations were redone without this variable, which allowed all observations to be included. To identify relationships between the various environmental factors and the relative abundance of the species, a direct gradient analysis was performed using stepwise canonical correspondence analysis (CCA; Ter Braak 1987) with the CANOCO software package. To do this, the relative basal area of each species was calculated as percentage of the basal area of all species present in the same plot. The forward selection is based on a Monte Carlo permutation test with 999 permutations; environmental variables were retained at a significance level of 0.001. To test if the selected variables play a significant role in determining the relative abundance of the species, an unconstrained analysis of the species data by correspondence analysis (CA) was performed. The resulting sample scores were compared with those obtained by CCA through the calculation of a correlation coefficient for each of the four ordination axes. Logistic regression and the gradient analyses were performed for the complete data set and for each geomorphological shore type separately. In the latter case, only those species present in at least 1 b plots of the corresponding shore type were considered.