Presettlement forest composition of the Superior-Martel Forest Prepared by: Fred Pinto, Stephen Romaniuk and Matt Ferguson Southern Science and Information Ministry of Natural Resources North Bay, Ontario Methods We conducted our study in the recently amalgamated J.E. Martel Forest and Superior Forest Management Units (hereafter referred to as the “Superior-Martel Forest”) in northeastern Ontario (Figure 1). The pre-settlement forest composition was reconstructed from Ontario Crown Land Survey Notes for the Superior-Martel Forest for the 1889-1923 period (Table 1). N W E S 100 0 100 200 Miles Figure 1. Outline of the township boundaries in the area that includes the Superior- Martel Forest in northeastern Ontario, Canada. The provinces defined in the inset map include Ontario (ON), Quebec (QC), and Manitoba (MB). Table 1. Summary of work done by Crown Land Surveyors in the Superior-Martel Forest. Surveyor Date(s) of survey Township boundaries A. Niven 1889 6 C.V. Gallagher 1923 17 Cavana & Watson 1901, 1903, 1919 28 G.B. Abrey 1902 2 H.J. Beatty 1905, 1906, 1921 11 J.M. Watson 1919 1 J.S. Dobie 1908, 1910, 1911, 1913 24 J.W. Fitzgerald 1920, 1921, 1923 33 J.W. Pearce 1912 1 L.V. Rorke 1905, 1906 13 Lang & Ross 1911 13 McAuslan, Anderson & Moore 1923 14 Speight & VanNostrand 1909, 1920, 1921 31 T.B. Speight 1898, 1899 6 T.J. Patten 1910, 1911 14 W. Beatty 1906, 1907 17 W. Galbraith 1901, 1902 2 Unknown Unknown 5 Total 1889-1923 238 The Ontario survey notes contain a wealth of information collected in the field, including a description of the tree cover along the boundaries of each township, the location and extent of forest stands along these boundaries, and a list of the tree species and genera present within each stand. Stands were delineated based on the changes in composition (proportion accounted for by each taxon) or changes in the order in which tree species were listed. The data in our study included only observations along the township boundaries and excluded data from lot and concession lines within the townships. Because full township surveys were not consistently completed across northeastern Ontario, the full survey data was not used. This approach also allowed us to include a larger, more consistent study area in our analysis. The forest resource inventory data that describe the current species composition of the Superior-Martel Forest are based on the interpretation of 1992 aerial photographs with interpretation completed in 1995, and updated with harvesting forecasts to 2001 and depletions to 1999. Also, “free to grow” regeneration (trees taller than the competing vegetation and at a height where survival is likely---usually about 1 m in Ontario) surveys were included in 1999. Comparison of FRI data along the township boundary to the abundance of tree species described by the FRI data within the township The survey data report only the species composition along township boundaries, and unlike the data in the provincial forest resource inventory, do not cover the entire area within a township. To determine whether species abundance along the boundary of a township adequately predicted the abundance of tree species within the whole township, we transcribed the inventory data along the township boundary lines and compared the results with inventory data for the whole township. The strength of the similarity determines the validity of using the historical survey notes to estimate the species composition of entire townships and thus, the validity of comparing the historical data with the current data as a means of examining post-settlement changes. Data from most (71%) of the townships in the Superior-Martel Forest were used in our analysis. Data from the other 29% of the townships were incomplete (these townships had Ontario Land Surveyor notes for less than three boundaries) and were therefore excluded from our analysis. The data from the forest resource inventory were mapped to the level of individual townships using the Arcview 3.2 GIS software (Environmental Systems Research Institute 1996), and the area of forested polygons was used to calculate the percentage composition of the first-listed species; this is the dominant species and (in most cases) the working group species for the township. The resulting dataset represented the percentage composition for the whole township. We ran an intersect operation in Arcview to retrieve the stand descriptions from the forest resource inventory along township boundaries using the Land Survey transects as a reference. This let us establish a correspondence between the inventory data and the survey data. We then compared these intersected boundaries with the percentage composition for the whole township to determine how accurately the boundary descriptions described the overall forest of the township. We used a paired t-test to compare the species compositions of each of the 87 townships after arcsine and square root transformations to improve normality. The results were pooled to obtain a single composition for each species. The data represent an unbiased sample of township boundaries because we included all the townships in the Superior-Martel Forest for which at least 3 boundaries were available. Most townships were 6 miles by 6 miles (10 X 10 km), but some were 9 miles by 9 miles (13.5 km by 13.5 km). The same boundary lines were included in the OLS and FRI comparisons described below. Comparison of species abundance between historical and current data Current (forest resource inventory) and historical (survey) data were compared along each township boundary using individual boundaries as the sampling unit (n=238), not townships. Analyses were performed at two levels. The first comparison included only the first-listed species in each stand, and summarized the length of the township boundary occupied by each species as a percentage of the total length of the boundary. Records of instructions given to land surveyors in the past suggest that tree species in a stand had to be listed in order of their abundance (Canada Department of Crown Lands 1862, 1867; Gentilecore and Donkin 1973). One aspect of the data supports this assumption, as sometimes adjacent stands had identical taxa, but their order differed, suggesting the use of a ranking system. The second comparison calculated an importance value called the “ranked abundance” for the historical and current data. The ranked abundance represents the length of the boundary occupied by each species, but weighted based on the rank of the species (the order in which it was listed) in a stand; the first species was given a weighting of three times its actual length, versus two times for the second species and once for the third and subsequent species. Since we are unsure if the instructions to list tree species by abundance were followed or followed by all surveyors, we also calculated an importance value called “equal importance” using the actual (unweighted) length for all species in each stand. The equal abundance value was calculated for both the historic and the current datasets. We used the ranked and equal importance values to calculate the percentage composition of each boundary being compared. Paired t-tests were run on the data after arcsine and square-root transformations. Statistical analyses were performed on the 238 boundaries, but the results presented have been pooled for the whole Superior- Martel Forest. When pine was mentioned in the historical data, we classified it as red pine and white pine; jack pine was always specified as jack pine, banksian pine, or pitch pine, and was thus not included in the “unspecified pine” or “total pine” groupings used in the analyses (Table 2). Additional first-listed species in the historical data that accounted for less than 1% of the forest composition and sample size were excluded from the analyses. These species included black ash (Fraxinus nigra Marsh.), balsam poplar (Populus balsamifera L.), ironwood (Ostrya virginiana (Mill.) K. Koch), willow (Salix spp.) and alder (Alnus spp.). The location of uncommon tree species recorded by the land surveyors are summarized in Table 3. Table 2. Common and scientific names associated with the names and codes used in the land survey notes. The last four rows represent the groupings we used in our analyses. Survey notes Interpretation Scientific name Black ash Black ash (AB) Fraxinus nigra Marsh. White ash White ash (AW) Fraxinus Americana L. Alder Alder spp. Alnus spp. Balsam Balsam fir (B) Abies balsamea (L.) Mill. Unspecified birch Birch spp.(BIR) Betula spp. Paper birch or white birch White birch (BW) Betula papyrifera Marsh. Yellow birch Yellow birch (BY) Betula alleghaniensis Britton. Cedar Eastern white cedar (CE) Thuja occidentalis L. Tamarack Tamarack (L) Larix laricina (Du Roi) K. Koch. Unspecified maple Maple species (M) Acer spp. Hard maple Sugar maple (MH) Acer saccharum Marsh. Soft maple Red (MR) or silver maple (MS) A. rubrum L. or A. saccharinum L. Unspecified pine Pine spp. (P), excl. P. banksiana Pinus spp. Red pine, Norway pine Red pine (PR) Pinus resinosa Ait. White pine White pine (PW) Pinus strobus L. Banksian, pitch, or jack pine Jack pine (PJ) Pinus banksiana Lamb. Poplar Poplar species (PO) Populus spp. Unspecified spruce Spruce species (SP) Picea spp. Black spruce Black spruce (SB) Picea mariana (Mill.) B.S.P. Red spruce Red spruce (SR) Picea rubens Sarg. White spruce White spruce (SW) Picea glauca (Moench) Voss. All birch categories listed above Total birch (BW, BY, BIR) Betula spp. All maple categories listed above Total maple (M, MR, MS, MH) Acer spp. All pine categories listed above Total pine (P, PR, PW) Pinus spp. All spruce categories listed above Total spruce (SP, SB, SW) Picea spp. The surveyors often described only the genus of the trees observed along the township boundaries; this was the case for spruce, pine, maple, and birch. We attempted to separate these genera using the logistic regression model described below. Assumptions were required to separate individual species for stands in the historical data which had been recorded as “birch”, “maple”, “pine”, or “spruce”. These groupings consist of white birch vs. yellow birch, hard maple vs. soft maple, white pine vs. red pine, and black spruce vs. white spruce; the spruce genus will be used for an example of the method used to distinguish species from the genera. To separate the two spruce species that could have been recorded under the “spruce” category, we developed a model using binary logistic regression (SPSS 2002) on the current (to 1999) data on the presence or absence of both species in the Superior-Martel Forest. We created a separate database for spruce stands in the current data using a rank column containing values of 0 (denoting black spruce) and 1 (denoting white spruce). If both species appeared in a stand, the stand with the higher ranking (i.e., the species closer to the beginning of the list of species in the survey data) was included and the other species was excluded. The frequency of the spruce species in the current data determined the cutoff value used in the model. If the model was significant (model χ2; SPSS 2002) and showed greater than 70% overall classification accuracy in predicting the species based on current species associations, then the species would be changed by running the model again, but on the historic data. The result of the model was transformed into a value between 0 and 1. The cut-off value would be used to divide the transformed values into individual species; i.e. all values above the cut-off would be one species, and those less than or equal to the cut-off value would be coded as the other species. Regardless of success of the model, all four generic groupings were also maintained for non-assumption based comparisons. In the analysis of categories for the trees based on dominance, the historical and current first-listed species were regrouped into intolerant hardwoods (poplar and white birch), mid-tolerant to tolerant hardwoods (ash, maple and yellow birch), and conifers (balsam fir, cedar, tamarack, pine, and spruce), and compared based on the proportion of a township boundary‟s total length occupied by each group. We used a paired t-test to compare the groups after arcsine and square root transformations to improve normality. The percent composition of taxa in the first three ranks was used to create a dominance index so as to detect any changes in species dominance over time. The most abundant species in each position was assigned a rank of one, the second species a rank of two, and so forth. The sum of the ranks from the past data was divided by the sum of the ranks from the present data to come up with the index value. A figure was produced for species compositions with or without logistic regression changes. Forest fragmentation was analyzed by summarizing the historical and current datasets based on the land cover type (e.g. forest vs. agricultural land). Some of these land cover types were not directly comparable because of inherent differences in the data and in the classification schemes used in the two types of survey. The Canadian Pacific Railway reached Chapleau in 1885, bringing with it a demand for jack pine for ties needed to expand the railway. Small local scale cutting by contractors was taking place as early as 1891. The Austin and Nicholson Lumber Company operated in the area from 1901 to 1921, booming in 1916 as the top producer of railway ties in the British Commonwealth with up to 200 000 ties per year (Conn 2003). They also harvested a significant amount of spruce and red and white pine mostly during the period of 1907 to the 1930s (Thorpe unpublished). Of those townships in which logging is known to have taken place, only 10 boundaries were surveyed after such logging. Therefore, these boundaries were removed from the dataset to maintain a more undisturbed forest condition. Logging took place by 1907 in the townships of Gallagher and Panet, and by 1922 in Pattinson and Floranna (Thorpe unpublished). In these cases, the land surveys for these townships occurred before logging took place and were therefore retained for the analyses. There is some evidence in the Ontario survey notes to indicate that logging was encountered during these surveys in the Superior-Martel Forest before 1923. About 1km of survey line in the townships of Collishaw and Stover contained survey entries in 1920 and 1921 stating “cut area”, “recently cut” and “partially lumbered”. These last three boundaries were included in the analysis since only one kilometer of forest was removed in these two townships. Results Table 3. Uncommon species mentioned in the land survey data in the Superior-Martel Forest. Species Township boundary Surveyor Date Black ash Bird East G.B. Abrey 1902 Deans West T.J. Patten 1911 Edighoffer West J.S. Dobie 1913 Singapore West J.W. Fitzgerald 1923 Tooms West Unknown ? White ash Deans West T.J. Patten 1911 Beech Bounsall East J.W. Fitzgerald 1923 A reasonable logistic regression model was obtained for birch, maple, pine and spruce (Table 4). However, only 0.4% of the birch codes in the FRI were yellow birch and since the Superior-Martel forest lies predominately in the boreal region, all unknown birch codes in the OLS were coded as white birch for all analyses instead of adopting the model (Table 5). Table 4. Percentage of OLS transect that refers to codes of each unspecified genus (e.g. BIR) versus the total for all species of that genus (e.g. BIR, BW, BY). This is shown individually for the first three species using 1) the length of forest cover (metres) for each genus; and 2) the frequency of codes used within each genus. The classification accuracy of the FRI logistic regression model is shown along with the model chi-square. % by length (metres) % by frequency accuracy of model summary self-validation (χ2 and p-value) test of FRI SPP1 SPP2 SPP3 SPP1 SPP2 SPP3 Birch 53.1 64.3 58.9 59.7 68.2 59.5 96.8% 616.7 p<0.01 Maple 50.5 51.3 86.0 54.8 34.1 46.8 76.4% 24.1 p<0.01 Pine 11.7 27.1 18.9 10.7 26.9 15.9 91.0% 247.0 p<0.01 Spruce 97.1 86.5 81.7 96.8 86.4 85.3 74.0% 10910.7 p<0.01 Table 5. Frequency of known FRI codes within the genus in question. Also shown, are the codes predicted by the logistic regression model or assumptions used to separate the four genera into species. Species Percent length in current FRI Percent length in OLS predicted by model SPP1 SPP2 SPP3 SPP1 SPP2 SPP3 White birch (BW) 28.56 20.48 15.98 14.03 20.45 30.15 Yellow birch (BY) 0.20 0.72 0.81 0.14 0.26 0.40 Hard maple (MH) 1.12 0.57 0.92 0.33 0.36 0.30 Soft maple (MS) 0.06 1.05 2.45 0.05 0.05 0.00 Red pine (PR) 0.02 0.07 0.10 0.24 0.38 0.27 White pine (PW) 0.87 0.82 1.99 3.37 0.84 2.12 Black spruce (SB) 25.88 20.92 15.31 29.48 19.25 7.93 White spruce (SW) 0.65 6.50 13.33 8.19 6.24 6.97 The historical (Ontario Land Survey) data provided an acceptable measure of the overall composition of the Superior-Martel Forest. That is, the forest cover recorded along the boundaries between townships predicted the overall forest composition with 95% confidence in all cases except for balsam fir(B), soft maple (MS), and white spruce (SW), which were significantly different (=0.05) (Table 6). This means that sampling along the township boundaries may be insufficient to detect changes in abundance for these species. Only the balsam fir is noteworthy since all maples and spruce were lumped together for the analyses. Table 6. Tree species composition of the current Superior-Martel Forest along township boundaries as compared to the total forest (based on first-listed species). Composition is distance based for the township lines and area based for the entire forest. Township lines 2001 FRI data Total forest 2001 FRI data B 0.90 0.90* BW 28.21 28.12 BY 0.14 0.15 Total birch 28.35 28.27 CE 4.71 4.89 LA 0.81 0.67 MH 1.08 0.89 MS 0.05 0.08* Total maple 1.13 0.97 PJ 19.12 19.63 PO 17.51 18.52 PR 0.02 0.02 PW 0.88 0.62 Total pine 0.90 0.64 SB 25.96 24.98 SW 0.60 0.53* Total spruce 26.56 25.51 * significant difference between FRI lines (township lines described in the 2001 FRI) and area within the township (2001 FRI) at the 95% confidence interval. ** significant difference between FRI lines (township lines described in the 2001 FRI) and area within the township (2001 FRI) at the 99% confidence interval. Table 7a. Land survey data is displayed to show changes in first-listed species composition. Since not all birch, maple, pine and spruce were recorded by species in the land surveys, all entries were lumped at the Genus level. Each value represents the mean for all the boundary lines. Species listed as “inconclusive” had an insufficient sample size and we are not able to state with certainty that the changes found along township boundaries reflect changes to the whole forest area (Table 6). OLS (1889-1923) FRI (2001) Change B 4.29 0.90** inconclusive BIRN 7.61 n.a. BWN 6.65 28.59 BYN 0.12 0.16 Total birch 14.38 28.75** increased CE 2.19 4.33** increased L 0.87 0.68 decreased MN 0.19 n.a. MHN 0.18 1.13 MSN --- 0.07 Total maple 0.37 1.19** increased PJ 25.58 19.15** decreased PO 9.10 17.58** increased PN 0.44 n.a. PRN 0.09 0.02 PWN 3.23 0.87 Total pine 3.77 0.89** decreased SPN 36.55 n.a. SBN 1.11 25.85 SWN --- 0.67 Total spruce 37.66 26.52** decreased Other A N 1.80 0.00 ** significant difference between OLS data and FRI township lines at the 99% confidence interval. A the „other‟ group included alder, black ash, balsam poplar, ironwood and willow. N not analyzed. Table 7b. Land survey data are displayed to show changes in first-listed species composition. OLS maple, pine and spruce species are those predicted by the logistic regression models. All unknown birch was assumed to be white birch. OLS (1889-1923) FRI (2001) Change BW 13.85 28.59** increased BY 0.14 0.16 not significant MH 0.33 1.13** increased MS 0.05 0.07 not significant PR 0.24 0.02 not significant PW 3.44 0.87** decreased SB 29.51 25.85* decreased SW 8.22 0.67** decreased * significant different between OLS data and FRI township lines at the 95% confidence interval. ** significant different between OLS data and FRI township lines at the 99% confidence interval. 100.00 Land surveys 90.00 (1889-1937) 80.00 FRI (2001) 70.00 length (%) 60.00 50.00 40.00 30.00 20.00 10.00 0.00 Intolerant hardw ood * Mid- to tolerant Conifer-dominated * hardw ood Figure 2. Tree functional categories based on first-listed species (n=238), all unknown birch in the OLS data was assumed to be white birch (intolerant hardwood). * denotes a statistically significant difference between land survey and FRI data at the 99% confidence level. Table 8a. Comparison by importance value between OLS and FRI data. Each value represents the mean for all the boundary lines. Since not all birch, maple, pine and spruce were recorded by species in the land surveys, all entries were lumped at the Genus level (shaded rows). Species Township boundaries OLS Township boundaries FRI Ranked abundance Equal abundance Ranked abundance Equal abundance ALN 1.63 2.05 n.a. n.a. B 10.32 12.22 7.45* 10.06 BIRN 9.79 10.56 n.a. n.a. BWN 6.48 6.54 19.91 16.88 BYN 0.17 0.18 0.45 0.54 Total birch 16.44 17.28 20.36** 17.42 CE 2.82 3.15 5.53** 5.82** L 2.90 3.59 2.20 2.74 MN 0.30 0.37 n.a. n.a. MHN 0.19 0.19 0.85 0.79 MSN 0.02 0.03 0.88 1.19 Total maple 0.51 0.6 1.73** 1.98** PJ 18.84 17.19 15.88** 14.78* PO 10.59 10.79 16.16** 15.28** PN 0.40 0.41 n.a. n.a. PRN 0.21 0.26 0.05 0.06 PWN 3.10 3.51 1.30 1.56 Total pine 3.72 4.19 1.35** 1.62** SPN 28.97 25.70 n.a. n.a. SBN 1.81 1.87 23.92 22.85 SWN 0.01 0.01 5.41 7.43 Total spruce 30.79 27.58 29.33 30.27** OtherN 1.20 1.37 0.03 0.04 * significance at 95% confidence interval between ranked OLS and FRI township boundaries and between equal OLS and FRI township boundaries. ** significance at 99% confidence interval between ranked OLS and FRI township boundaries and between equal OLS and FRI township boundaries. N species not analyzed. Table 8b. Comparison by importance value between OLS data from 1889-1923 and 2001 updated FRI. Each value represents the mean for all the boundary lines. OLS maple, pine and spruce species are those predicted by the logistic regression models. All unknown birch was assumed to be white birch. Species Township boundaries OLS Township boundaries FRI Ranked abundance Equal abundance Ranked abundance Equal abundance BW 16.21 17.12 19.91** 16.88 BY 0.19 0.21 0.45** 0.54** MH 0.49 0.57 0.85** 0.79* MS 0.07 0.08 0.88** 1.19 PR 0.38 0.43 0.05** 0.06** PW 3.30 3.74 1.30** 1.56** SB 24.79 22.47 23.92 22.85 SW 6.04 5.19 5.41* 7.43** * significance at 95% confidence interval between ranked OLS and FRI township boundaries and between equal OLS and FRI township boundaries. ** significance at 99% confidence interval between ranked OLS and FRI township boundaries and between equal OLS and FRI township boundaries. The change in „forest‟ cover illustrated in Table 9 is misleading; the difference can be attributed to the differences in the classification used. That is, the historic non-productive forest, areas of burn (burned clean) and windfall, and sections of no data can be classified as productive forest by present standards. If we examine the 78.4% forest cover in the FRI data further to look at the breakdown of land ownership (Table 9), it is evident that 3.76% of this forest cover is classified as patent (or private) land. The majority of this patent land consists of mining claims that have likely been converted from forest cover to non-forest cover. Table 9. Proportion of total area in each land cover type based on the Ontario Land Survey (OLS) and Forest Resource Inventory (FRI) data. “Rock” refers to forested or non-forested land containing exposed bedrock. “n.a.” means “not applicable” (i.e., no land was classified in this category in the OLS or FRI data). Barren and scattered or not satisfactorily regenerated (NSR) can include areas that have been cutover and/or burned and may be comparable to the area burned that was specified in the OLS. Windthrow includes any historic transects that mentioned windthrow. Land cover type Percent of total area OLS FRI Forest 70.80 78.40 Non-productive forest (muskeg, alder) 12.40 4.56 Non-productive forest (rock) 0.08 0.03 Grass and meadows 0.30 --- Unclassified land 0.01 0.14 Water 7.45 7.71 Burned areas (burned clean) 4.17 --- Partial burns (over & above that burned clean) 9.72 --- Windthrow 0.44 --- Barren & scattered or NSR --- 4.75 No data 4.33 4.35 30 Maple 25 Pine 20 CE L Past rank 15 PO PJ 10 B Birch 5 Spruce 0 0 5 10 15 20 25 30 Present rank Figure 3a. Changes in dominance between the historic (past rank) and current (present rank) species groups. The 45° line represents a one to one correlation between the two ranks. Species groups above the line display an increase in dominance. Species groups with a smaller rank illustrate a higher dominance in forest stands. Each combined rank includes the compositions from each of the first three species positions. 40 MS 35 MH BY PR 30 25 Past rank PW CE 20 L SW 15 PO 10 SB B & PJ BW 5 0 0 5 10 15 20 25 30 35 40 Present rank Figure 3b. Changes in dominance between the historic (past rank) and current (present rank) individual tree species. The 45° line represents a one to one correlation between the two ranks. Species or species groups above the line display an increase in dominance. Species or species groups with a smaller rank illustrate a higher dominance in forest stands. Each combined rank includes the compositions from each of the first three species positions. The maple, pine and spruce species were those determined by the logistic regression models, all unknown birch was coded as white birch. Literature cited Canada Department of Crown Lands. 1862. Remarks on Upper Canada surveys, and extracts from the surveyor‟s reports, containing a description of the soil and timber of the townships in the Huron and Ottawa territory, and on the north shores of Lake Huron. Appendix no. 26, to the report of the Commissioner of Crown Lands, for 1861. Hunter, Rose, and Lemieux, Quebec. Canada Department of Crown Lands. 1867. Remarks on Upper Canada surveys, and extracts from the surveyor‟s reports, containing a description of the soil and timber of the townships in the Ottawa River and Georgian Bay Section and between the Spanish River, on the north shore of Lake Huron, and Goulay‟s Bay, on Lake Superior. Hunter and Rose, Ottawa, Ontario. Conn, Heather, The human history of Wakami Lake, 2003. http://www.canadianfishing.com/sultan/wakami.htm (2003, Dec 2). Environmental Systems Research Institute (ESRI). 1996. Arcview 3.2 GIS software. ESRI Inc., Redlands, California. 350pp. Gentilecore, L. and K. Donkin. 1973. Land surveys of southern Ontario. An introduction and index to the field notebooks of the Ontario Land Surveyors 1784-1859. Supplement #2 to Canadian Cartographer Vol. 10, 166pp. SPSS. 2002. Version - base 11.5, with Regression Models. SPSS Inc., Chicago, Illinois.
Pages to are hidden for
"Table 3"Please download to view full document