Review of Apparent Vertical Movement Rates in the Great Lakes Region Jacob Bruxer Chuck Southam I. Table of Contents II. List of Figures ........................................................................................................... iv III. List of Tables ............................................................................................................ vi 1. 2. 3. Introduction................................................................................................................ 1 Past Studies ................................................................................................................ 2 Method ....................................................................................................................... 2 3.1. Water Level Difference Plots and Linear Regression.......................................... 2 3.2. Variability ............................................................................................................ 3 3.3. Outliers................................................................................................................. 5 3.4. Running Average Plots ........................................................................................ 6 3.5. Breakpoint Analysis............................................................................................. 7 4. Lake Superior............................................................................................................. 8 4.1. Overview.............................................................................................................. 8 4.2. Point Iroquois, Michigan ..................................................................................... 9 4.3. Marquette, Michigan.......................................................................................... 10 4.4. Ontonagon, Michigan ........................................................................................ 10 4.5. Duluth, Minnesota.............................................................................................. 11 4.6. Two Harbors, Minnesota ................................................................................... 11 4.7. Grand Marais, Minnesota................................................................................... 12 4.8. Thunder Bay, Ontario ........................................................................................ 13 4.9. Rossport, Ontario ............................................................................................... 14 4.10. Michipicoten, Ontario ........................................................................................ 15 4.11. Gros Cap, Ontario .............................................................................................. 16 4.12. Summary ............................................................................................................ 17 5. Lakes Michigan-Huron ............................................................................................ 18 5.1. Overview............................................................................................................ 18 5.2. Harbor Beach, Michigan.................................................................................... 20 5.3. Ludington, Michigan.......................................................................................... 21 5.4. Holland, Michigan ............................................................................................. 22 5.5. Calumet Harbor, Illinois .................................................................................... 23 5.6. Milwaukee, Wisconsin....................................................................................... 25 5.7. Kewaunee, Wisconsin........................................................................................ 27 5.8. Sturgeon Bay Canal, Wisconsin ........................................................................ 28 5.9. Green Bay, Wisconsin ....................................................................................... 29 5.10. Port Inland, Michigan ........................................................................................ 29 5.11. Lakeport, Michigan............................................................................................ 30 5.12. Essexville, Michigan.......................................................................................... 31 5.13. Harrisville, Michigan ......................................................................................... 32 5.14. Mackinaw City, Michigan ................................................................................. 33 5.15. De Tour, Michigan............................................................................................. 35 5.16. Thessalon, Ontario ............................................................................................. 35 5.17. Little Current, Ontario ....................................................................................... 37 5.18. Parry Sound, Ontario ......................................................................................... 38 ii 5.19. Collingwood, Ontario ........................................................................................ 39 5.20. Tobermory, Ontario ........................................................................................... 41 5.21. Goderich, Ontario .............................................................................................. 42 5.22. Summary ............................................................................................................ 46 5.23. Outlet at Lakeport .............................................................................................. 47 6. Lake Erie.................................................................................................................. 50 6.1. Overview............................................................................................................ 50 6.2. Buffalo, New York............................................................................................. 51 6.3. Sturgeon Point, New York................................................................................. 51 6.4. Barcelona, New York......................................................................................... 52 6.5. Erie, Pennsylvania.............................................................................................. 52 6.6. Fairport, Ohio..................................................................................................... 53 6.7. Cleveland, Ohio ................................................................................................. 54 6.8. Marblehead, Ohio .............................................................................................. 56 6.9. Toledo, Ohio ...................................................................................................... 57 6.10. Monroe, Michigan.............................................................................................. 60 6.11. Fermi Power Plant, Michigan ............................................................................ 61 6.12. Bar Point, Ontario .............................................................................................. 62 6.13. Kingsville, Ontario............................................................................................. 63 6.14. Erieau, Ontario................................................................................................... 64 6.15. Port Stanley, Ontario.......................................................................................... 65 6.16. Port Dover, Ontario............................................................................................ 66 6.17. Port Colborne, Ontario....................................................................................... 67 6.18. Summary ............................................................................................................ 70 7. Lake Ontario ............................................................................................................ 71 7.1. Overview............................................................................................................ 71 7.2. Cape Vincent, New York................................................................................... 72 7.3. Oswego, New York............................................................................................ 73 7.4. Rochester, New York......................................................................................... 74 7.5. Olcott, New York............................................................................................... 74 7.6. Port Weller, Ontario........................................................................................... 75 7.7. Port Dalhousie, Ontario ..................................................................................... 76 7.8. Burlington, Ontario ............................................................................................ 77 7.9. Toronto, Ontario ................................................................................................ 78 7.10. Cobourg, Ontario ............................................................................................... 80 7.11. Kingston, Ontario............................................................................................... 81 7.12. Summary ............................................................................................................ 83 8. Conclusions.............................................................................................................. 84 9. Recommendations.................................................................................................... 86 10. References................................................................................................................ 88 iii II. List of Figures Figure 3-1: Water level difference plot of Point Iroquois minus Duluth (Lake Superior). ............ 3 Figure 3-2: Harbor Beach minus Calumet Harbor water level difference plot with 1907 outlier included. .......................................................................................................................................... 6 Figure 3-3: Water level difference plot of Harbor Beach minus Milwaukee for the period 1904 to 2006 with 5-year centred runing average. ....................................................................................... 7 Figure 3-4: Harbor Beach minus Goderich water level difference plot, with piecewise regression and the estimated breakpoint indicated. .......................................................................................... 8 Figure 4-1: Lake Superior gauge locations...................................................................................... 9 Figure 4-2: Point Iroquois minus Marquette water level difference plot. .................................... 10 Figure 4-3: Point Iroquois minus Ontonagon water level difference plot. ................................... 11 Figure 4-4: Point Iroquois minus Two Harbors water level difference plot.................................. 12 Figure 4-5: Point Iroquois minus Grand Marais water level difference plot................................. 13 Figure 4-6: Point Iroquois minus Thunder Bay water level difference plot.................................. 14 Figure 4-7: Point Iroquois minus Rossport water level difference plot......................................... 15 Figure 4-8: Point Iroquois minus Michipicoten water level difference plot. ................................ 16 Figure 4-9: Point Iroquois minus Gros Cap water level difference plot........................................ 17 Figure 5-1: Lakes Michigan-Huron gauge locations..................................................................... 19 Figure 5-2: Harbor Beach minus Ludington water level difference plot....................................... 21 Figure 5-3: Milwaukee minus Ludington water level difference plot........................................... 22 Figure 5-4: Harbor Beach minus Holland water level difference plot. ......................................... 23 Figure 5-5: Harbor Beach minus Calumet Harbor water level difference plot with 1907 outlier removed. ........................................................................................................................................ 24 Figure 5-6: Harbor Beach minus Calumet Harbor water level difference plot with 5-year centred running average. ............................................................................................................................ 24 Figure 5-7: Harbor Beach minus Milwaukee water level difference plot. .................................... 25 Figure 5-8: Harbor Beach minus various Lake Michigan gauges 5-year centred running averages of water level difference plots. ...................................................................................................... 26 Figure 5-9: Harbor Beach minus Kewaunee water level difference plot. ..................................... 27 Figure 5-10: Harbor Beach minus Sturgeon Bay Canal water level difference plot. .................... 28 Figure 5-11: Harbor Beach minus Green Bay water level difference plot. ................................... 29 Figure 5-12: Harbor Beach minus Port Inland water level difference plot. .................................. 30 Figure 5-13: Harbor Beach minus Lakeport water level difference plot....................................... 31 Figure 5-14: Harbor Beach minus Essexville water level difference plot..................................... 32 Figure 5-15: Harbor Beach minus Harrisville water level difference plot. ................................... 33 Figure 5-16: Harbor Beach minus Mackinaw City water level difference plot............................. 34 Figure 5-17: Harbor Beach minus Mackinaw City water level difference plot with 5-year centred running average. ............................................................................................................................ 34 Figure 5-18: Harbor Beach minus De Tour water level difference plot........................................ 35 Figure 5-19: Harbor Beach minus Thessalon water level difference plot. .................................... 36 Figure 5-20: Harbor Beach minus Thessalon water level difference plot with 5-year centred running average. ............................................................................................................................ 37 Figure 5-21: Harbor Beach minus Little Current water level difference plot. .............................. 38 Figure 5-22: Harbor Beach minus Parry Sound water level difference plot. ................................ 39 Figure 5-23: Harbor Beach minus Collingwood water level difference plot. ............................... 40 Figure 5-24: Harbor Beach minus Collingwood water level difference plot with 5-year centred running average. ............................................................................................................................ 40 Figure 5-25: Harbor Beach minus Collingwood breakpoint estimation........................................ 41 iv Figure 5-26: Harbor Beach minus Tobermory water level difference plot. .................................. 42 Figure 5-27: Harbor Beach minus Goderich water level difference plot. ..................................... 43 Figure 5-28: Goderich minus Thessalon water level difference plot. ........................................... 44 Figure 5-29: Goderich minus Collingwood water level difference plot........................................ 45 Figure 5-30: Goderich minus Milwaukee water level difference plot........................................... 46 Figure 6-1: Lake Erie gauge locations........................................................................................... 50 Figure 6-2: Buffalo minus Sturgeon Point water level difference plot. ........................................ 51 Figure 6-3: Buffalo minus Barcelona water level difference plot. ................................................ 52 Figure 6-4: Buffalo minus Erie, PA, water level difference plot................................................... 53 Figure 6-5: Buffalo minus Fairport water level difference plot. ................................................... 54 Figure 6-6: Buffalo minus Cleveland water level difference plot. ................................................ 55 Figure 6-7: Buffalo minus Cleveland water level difference plot with 5-year centred running average........................................................................................................................................... 56 Figure 6-8: Buffalo minus Marblehead water level difference plot. ............................................. 57 Figure 6-9: Buffalo minus Toledo water level difference plot. ..................................................... 58 Figure 6-10: Comparison of Toledo minus Fermi Power Plant and Toledo minus Erie, PA water level difference plots. .................................................................................................................... 59 Figure 6-11: Buffalo minus Toledo breakpoint estimation for recording gauge period................ 60 Figure 6-12: Buffalo minus Monroe water level difference plot................................................... 61 Figure 6-13: Buffalo minus Fermi Power Plant water level difference plot. ................................ 62 Figure 6-14: Buffalo minus Bar Point water level difference plot. ............................................... 63 Figure 6-15: Buffalo minus Kingsville water level difference plot............................................... 64 Figure 6-16: Buffalo minus Erieau water level difference plot..................................................... 65 Figure 6-17: Buffalo minus Port Stanley water level difference plot............................................ 66 Figure 6-18: Buffalo minus Port Dover water level difference plot.............................................. 67 Figure 6-19: Buffalo minus Port Colborne water level difference plot......................................... 68 Figure 6-20: Buffalo minus Port Colborne water level difference plot with 5-year centred running average........................................................................................................................................... 68 Figure 6-21: Buffalo minus various Lake Erie gauges 5-year centred running averages of water level difference plots. .................................................................................................................... 69 Figure 6-22: Comparison of Buffalo minus Port Dover and Buffalo minus Port Colborne water level difference plots. .................................................................................................................... 70 Figure 7-1: Lake Ontario gauge locations. .................................................................................... 72 Figure 7-2: Cape Vincent minus Oswego water level difference plot........................................... 73 Figure 7-3: Cape Vincent minus Rochester water level difference plot........................................ 74 Figure 7-4: Cape Vincent minus Olcott water level difference plot.............................................. 75 Figure 7-5: Cape Vincent minus Port Weller water level difference plot. .................................... 76 Figure 7-6: Cape Vincent minus Port Dalhousie water level difference plot................................ 77 Figure 7-7: Cape Vincent minus Burlington water level difference plot. ..................................... 78 Figure 7-8: Cape Vincent minus Toronto water level difference plot........................................... 79 Figure 7-9: Cape Vincent minus Toronto water level difference plot with 5-year centred running average........................................................................................................................................... 79 Figure 7-10: Cape Vincent minus Cobourg water level difference plot (June to Sep. average). .. 80 Figure 7-11: Cape Vincent minus Cobourg water level difference plot (July to Oct. average). ... 81 Figure 7-12: Cape Vincent minus Kingston water level difference plot. ...................................... 82 Figure 7-13: Comparison of Kingston water level difference plots. ............................................. 83 v III. List of Tables Table 4-1: Lake Superior gauge stations and periods of record. ..................................................... 9 Table 4-2: Summary of Lake Superior rates of apparent vertical movement and uncertainty relative to Point Iroquois. .............................................................................................................. 18 Table 5-1: Lake Michigan gauge stations and periods of record................................................... 19 Table 5-2: Lake Huron gauge stations and periods of record........................................................ 20 Table 5-3: Summary of Lakes Michigan-Huron rates of apparent vertical movement and uncertainty relative to Harbor Beach............................................................................................. 47 Table 5-4: Comparison of rates of movement on Lakes Michigan-Huron relative to Harbor Beach and Lakeport.................................................................................................................................. 48 Table 5-5: Summary of Lakes Michigan-Huron rates of apparent vertical movement and uncertainty relative to Lakeport..................................................................................................... 49 Table 6-1: Lake Erie gauge stations and periods of record. .......................................................... 50 Table 6-2: Summary of Lake Erie rates of apparent vertical movement and uncertainty relative to Buffalo........................................................................................................................................... 71 Table 7-1: Lake Ontario gauge stations and periods of record...................................................... 72 Table 7-2: Summary of Lake Ontario rates of apparent vertical movement and uncertainty relative to Cape Vincent. ............................................................................................................................ 84 vi vii 1. Introduction During the last glacial era, glaciers covered much of North America, including the Great Lakes region. Under the great weight of these glaciers the Earth’s crust was deformed. When the ice in the Great Lakes region melted some 10,000 years ago, the crust began to rebound, and it is believed that this rebounding has continued ever since. As the glaciers receded, they generally tended to do so in a north-easterly direction. For this reason, the northern and eastern areas of the Great Lakes regions tended to be covered with a greater amount of ice for a longer period of time than areas to the south and west. As such, rebound rates have been observed to occur at a faster rate as one travels to the north and to the east. In general, the result is that observed water levels on a given lake appear to be falling over time on the northern and eastern shores, while water levels appear to be rising over time on the southern and western shores. By comparing the water level differences over time between pairs of gauges on a given lake, the relative rate of apparent vertical movement between these gauges can be estimated. The phrase “apparent vertical movement” has been used intentionally here in place of “glacial isostatic adjustment”, “post-glacial rebound”, “crustal movement”, or any other phrase that would suggest movement of the Earth’s crust specifically. While often the apparent vertical movement between gauges is assumed to be exclusively the result of glacial isostatic adjustment, in some cases the movement may be the result of other known or unknown factors. An example would be at a gauge station where in the immediate local vicinity the land, including both the gauge and the local benchmark, is subsiding as a result of something other than glacial isostatic adjustment. In fact such an occurrence has been observed on Lake Erie at the Fairport, Ohio gauge. Other such events that could be responsible for inconsistencies in the water level difference plots may include the relocation of a gauge or a change in the data collection method from staff gauge readings to continuously recording water level measurements. On the other hand, jumps, shifts, changes in rate, step-like patterns or other discontinuities in water level difference plots might actually be the result of a poorly understood glacial isostatic adjustment process. For example, the assumption of a linear trend over the short-term between gauge pairs may be incorrect. Linear trend lines calculated from water level difference plots are used to determine the rate of apparent vertical movement between gauges; however, in reality the processes responsible for observed glacial isostatic adjustment may be non-linear, but may instead follow some different pattern or no pattern at all. The use of long-term linear regression of water level differences may be useful for some purposes, but not all. For these reasons, this report provides a revision of the rates of glacial isostatic adjustment as calculated from water level difference plots using linear regression, and also illustrates the difficulties and anomalies that occur in this type of analysis. Significant discontinuities in the water level difference plots are identified, and where possible, potential causes for the discontinuities are discussed. 1 2. Past Studies Glacial isostatic adjustment has been studied for some time in the Great Lakes region. A report by the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data (Coordinating Committee, 1977) summarizes a number of earlier studies. The Coordinating Committee compared differences between the four-month mean June to September monthly average water levels for gauge pairs on each of the Great Lakes. Linear regression lines were fit to the water level difference versus time plots, with the slope of the line indicating the relative rate of movement between the two gauges in question. Other researchers (e.g. Tushingham, 1992) have used similar methods that are referenced in the Coordinating Committee’s 2001 report which provides one of the most recent updates. In their study, the Coordinating Committee used monthly mean water level data from all twelve months with outliers removed, as opposed to the traditional June to September average used in earlier studies. The vertical velocity rates and standard errors were calculated for all gauges from this data. The findings were also reported in Mainville and Craymer (2005). It should be noted that these past studies did not deal directly with non-homogeneous data, trend shifts, step-like patterns, or other inconsistencies, whereas this current study does. 3. Method 3.1. Water Level Difference Plots and Linear Regression Similar to past studies, plots of water level differences versus time between gauge pairs on each lake were constructed, and a linear least-squares regression line was fit to the data. The slope of this line indicates the approximate rate of apparent vertical movement between the two gauges. Published monthly mean water level data at gauges across the Great Lakes region for the period of record up until 2006 was obtained for use in this study. Data for Canadian gauges was obtained from the Canadian Hydrographic Service, while data from U.S. gauges was obtained from the U.S. National Ocean Service. As an example, Figure 3-1 shows a water level difference plot for Point Iroquois and Duluth. Point Iroquois is located at the eastern end and approximate outlet of Lake Superior, while Duluth is located at the western end of the lake. Rates of apparent vertical movement are normally reported relative to the outlet of the lake, but can also be determined for any gauge pair on a given lake. The four-month (June to September) mean monthly water level as recorded at Duluth is subtracted from the same four-month mean level as recorded at Point Iroquois for all years of record and plotted as shown. The 2 downward slope of the linear regression line indicates that over time the water level observed at Duluth appears to be increasing with respect to the water level recorded at Point Iroquois at a rate of approximately 0.0027 metres/year (m/year) or 27 centrimetres/century (cm/century). This apparent vertical movement is normally assumed to be the result of the earth’s crust at Duluth falling with respect to the crust at Point Iroquois. Please note that in absolute terms (i.e. relative to the earth’s geo-centre) the crust could be falling or rising at either or both gauge locations, but that it is not possible to determine what the case is from these diagrams. Rather it is the relative rate between the two gauges that is illustrated by the water level difference plots. In addition to the water level difference plots and calculated rates of vertical movement, the uncertainty as given by the standard deviation of the slope value was also calculated. Point Iroquois - Duluth 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1920 1940 1960 1980 2000 2020 Year Figure 3-1: Water level difference plot of Point Iroquois minus Duluth (Lake Superior). Water levels were averaged from June to September in this case. 3.2. Variability Figure 3-1 also illustrates the fact that plots of water level differences versus time show noise or variability in addition to the assumed linear trend. This is believed to be primarily the result of meteorological impacts, such as storm and wind events, differences in barometric pressure, ice effects at gauge locations, etc. To help reduce some of this 3 variability, in past studies monthly mean water level differences between gauge pairs have normally been averaged. A four-month average from June to September was commonly used as these four months were believed to be those least susceptible to the meteorological impacts outlined. One notable exception was the recent Coordinating Committee (2001) report, which used monthly values from all twelve months as opposed to the four-month June to September average approach. Outliers were removed in this study where appropriate, and while the plots using data from all months tend to show greater variability than fourmonth average plots, the linear trend in the data can still be identified. However, one of the problems that results from using this method is that the density of the data and the greater variability can mask important features in the water level differences plots, such as non-linear patterns, shifts, steps and other discontinuities. Taking this into consideration, initial investigations conducted for this study tried to identify which months, if any, showed the least amount of noise and variability with regards to water level differences between gauge pairs. It was believed that these months would be the best choices for use in the average water level difference plots used to infer apparent vertical movement of water level gauges over time. Each lake was investigated individually in this analysis, since the lakes vary significantly in their geographic and climatic attributes, which could cause certain months of data to be most appropriate for use on a certain lake but not on others. For example, water levels on Lake Erie, due to such factors as the lake’s size, depth, geographic orientation and climate, tend to be more affected by wind setup and seiche effects than some of the other lakes. Therefore, the best months used on Lake Erie (i.e. those least affected by meteorological impacts) may not be the most ideal months for use on the other lakes. A number of tests were done on water level difference plots for each lake and for each month. These included determining the mean and standard deviations of the monthly differences, determining the standard deviation of the year-to-year lag of the monthly differences, and calculating the R-squared value of the monthly difference plots over time. The six traditionally ice-free months of May to October generally performed better than the other months in these tests, which was as expected. Between these six months, however, the results did not show conclusive evidence that any months in particular would be significantly better for use in the water level difference analysis than others. The latter months (July to October) showed the least variability in terms of mean differences, while earlier months (May to August) showed the lowest variability when standard deviations about the mean year-to-year differences were plotted. These months also had the best fit according to the R-squared statistic. July and August tended to be the best regardless of test chosen, but again, these did not perform significantly better than the other ice-free months investigated. Somewhat surprisingly, this was the case regardless of the lake chosen. 4 For a number of reasons, it was decided that the traditional four-month, June to September average be used primarily in this study. First, as noted, none of the months performed especially better than any of the other months, but the months of June to September were four of the best according to the variability investigation. Second, the uncertainty is reduced in linear regression calculations as more data is available, which includes both the number of years of available data, as well as the completeness of these available years. Water level records are sometimes missing in the period of record for a given gauge due to equipment malfunctions, gauge maintenance, relocation or any number of other reasons. Since all months of data for a given average calculation must be available for the average to be computed and used in the regression analysis, the more months required for a calculation, the more likely it is that one or more of these months are missing. If one or more months are missing, the average cannot be calculated and this year will have to be omitted. For example, if only the month of February is missing in a given year, then the annual average for that year cannot be calculated and the year cannot be used in the regression analysis. On the other hand, the June to September average for this same year can be calculated and used. As a result, averages that include a smaller number of months (such as the four-month June to September average) can be calculated more often than averages that require more months (such as the 12-month annual average or even a six-month May to October average). An alternative option would be to use less than four months, or even choose a single month of data as opposed to an average calculation; however, the use of an average helps to reduce some of the variability and noise in the water level difference plots. The choice of a four-month average (in this case June to September) was done in part to balance the need for available data and the need to reduce variability in the water level difference relationship. Lastly, using the June to September average is consistent with many previous studies, thus allowing for direct comparison. In addition to those studies already mentioned, June to September averages have also been used for International Great Lakes Datum (IGLD) adjustments. Furthermore, rates of apparent vertical movement were calculated during this study for other combinations of months, and in general the results normally tended to be in close agreement. Additional calculated rates are reported in this document where deemed necessary. However, unless otherwise stated, the rates of apparent vertical movement determined from the four-month June to September average water level differences are reported. 3.3. Outliers Outliers were subjectively removed only if they were obviously the result of some issue beyond those known to cause normal variability observed in the water level difference plots, or if they had a significantly disproportional impact on the calculated rate of apparent vertical movement. Only a small number of outliers were removed in this way, and those removed are documented in this report. It should be noted that normally even these most extreme outliers did not make a significant difference in the calculated rates of vertical movement. As an example, Figure 3-2 shows one of the most 5 significant outliers removed, occurring in 1907 in a plot of Harbor Beach minus Calumet Harbor. When this outlier was removed the difference in calculated rate of apparent vertical movement was only 0.80 cm/century. Harbor Beach - Calumet Harbor 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 3-2: Harbor Beach minus Calumet Harbor water level difference plot with 1907 outlier included. 3.4. Running Average Plots In order to smooth out some of the noise in the water level difference plots, running averages of the water level differences were calculated for certain gauge pairs. Three- to nine-year centred averages were calculated about the central year and plotted against the water level differences. By smoothing out some of the noise in the water level difference plots, the running average plots are able to better illustrate potential shifts, changes in slope, stepwise patterns, or other features that may not be as apparent in the original water level difference plots. An example of a running average plot is shown in Figure 3-3. The June to September average differences are plotted along with the five-year centred running averages for June to September. While also evident in the original water level difference plot, the running average plot more clearly indicates what appears to be a step-like pattern in the calculated differences. 6 Harbor Beach - Milwaukee 0.3 June-Sep. Avg. June-Sep. 5-Year Centred Avg. Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 3-3: Water level difference plot of Harbor Beach minus Milwaukee for the period 1904 to 2006. The June to September average differences are plotted along with the June to September five-year centred running average plot showing a step-like pattern. 3.5. Breakpoint Analysis Breakpoint analysis was performed on certain water level difference plots if evidence suggested a distinct change in rate of apparent vertical movement has occurred. For example, Figure 3-4 shows a plot of Harbor Beach minus Goderich June to September average water level differences. Since visual inspection of the original plot suggested a change in slope at some point around the 1950’s, piecewise regression was performed and the estimated breakpoint identified. Linear regression is performed before and after the breakpoint, which is assumed to exist. The slope and intercept of lines before and after the breakpoint are calculated, and the breakpoint itself is identified as the year in which the overall R-squared value of the fitted lines is a maximum. This was found using the Solver function in Microsoft Excel. In the case of Harbor Beach minus Goderich, the breakpoint was identified at 1955. Due to the uncertainty and variability in the water level difference plots, the breakpoint found in this way can only be accepted as an estimate. Nonetheless, this method gives an unbiased means of estimating the breakpoint, and also allows subjective comparisons to be made between different plots. 7 Harbor Beach - Goderich 0.3 June-Sep. Avg. Piecewise Regression Estimated Breakpoint Water Level Difference (m) 0.2 0.1 1955 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 3-4: Harbor Beach minus Goderich water level difference plot, with piecewise regression and the estimated breakpoint indicated. 4. Lake Superior 4.1. Overview Water level data from a total of ten gauge stations were analyzed on Lake Superior (Figure 4-1). Four of these gauge stations are located in Canada, while the remaining six gauge stations are located in the United States. Table 4-1 gives a list of the gauge stations used in this analysis, as well as a summary of the periods of record for each station. Data at some locations were divided between gauges due to gauge relocations. These datasets were combined and treated as one homogeneous dataset in this analysis unless otherwise noted, and are identified in the Station ID with a backslash indicating the two final digits of the two gauges (for example, Marquette data comes from two individual gauges, 9099016 and 9099018, which cover different time periods at essentially the same location). Furthermore, if a full year at a given station contained no data (i.e. no monthly data available from January to December for a given year) this year was identified. Only full years of missing data are identified, but note also that additional years within the dataset not identified may still be missing individual or many months. This applies to gauges on all of the Great Lakes. 8 Figure 4-1: Lake Superior gauge locations. Start End Missing (Full) Location Year Year Years Point Iroquois, MI, USA 9099004 1930 2006 1945-49 9099016/8 Marquette, MI, USA 1860 2006 -9099044 Ontonagon, MI, USA 1959 2006 -9099064 Duluth, MN, USA 1860 2006 1874-79 Two Harbors, MN, 1888-98; 1901-29; USA 1932-34; 1936-40 9099070 1887 1988 Grand Marais, MN, 9099090 1966 2006 -USA Thunder Bay, ON, 10050 1907 2006 1912-1913 Canada Rossport, ON, Canada 10220 1967 2006 -Michipicoten, ON, 10750 1915 2006 -Canada Gros Cap, ON, Canada 10920 1926 2006 1930-1960 Table 4-1: Lake Superior gauge stations and periods of record. Station ID 4.2. Number of Full Years Missing Total Years of Data 5 0 0 6 72 147 48 141 48 54 0 41 2 98 0 40 0 92 31 50 Point Iroquois, Michigan The Point Iroquois gauge was selected as the Lake Superior outlet. Gros Cap is also located close to the outlet of Lake Superior; however, Point Iroquois has traditionally been the gauge used to represent the outlet, likely owing to the fact that it has a longer period of record than Gros Cap. Data at Point Iroquois begins in 1930 and continues through to 2006, with the exception of a period of missing data from October 1944 through October 1950. The gauge was not moved during this period of missing data, but it was moved in 1970. All data at Point Iroquois was obtained from a continuous recording gauge. 9 4.3. Marquette, Michigan The water level record at Marquette is a combination of data from two gauge stations. The data begins in 1860 and was collected at gauge 909-9016 (“Marquette”) until 1980. The data collected from 1980 until 2006 comes from gauge 909-9018 (“Marquette C.G.”). The data were merged, with data from station 909-9018 being used for the overlapping year of 1980, and it was assumed that this represented a homogeneous dataset. The water level difference plot for Point Iroquois minus Marquette is shown in Figure 4-2. The rate of vertical movement as determined from the slope of the linear regression line is -10.50 cm/century, indicating that Marquette is falling relative to the lake’s outlet at Point Iroquois. Visual inspection of the plot does not reveal any significant discontinuities in the slope of the linear relationship. Point Iroquois - Marquette 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0.0 -0.1 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 Year Figure 4-2: Point Iroquois minus Marquette water level difference plot. 4.4. Ontonagon, Michigan The water level record at Ontonagon begins during 1959, with the first June to September average being calculated in 1960. The water level difference plot for Point Iroquois minus Ontonagon is shown in Figure 4-3. The rate of vertical movement as determined from the slope of the linear regression line is -17.32 cm/century, indicating 10 that Ontonagon is falling relative to the lake’s outlet. Visual inspection of the plot does not reveal any significant discontinuities. Point Iroquois - Ontonogan 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 4-3: Point Iroquois minus Ontonagon water level difference plot. 4.5. Duluth, Minnesota The water level record begins in 1860 at Duluth. The water level difference plot for Point Iroquois minus Duluth was shown previously in Figure 3-1. The rate of vertical movement as determined from the slope of the linear regression line is -26.85 cm/century, indicating that Duluth is falling relative to the lake’s outlet. Data at Duluth prior to 1950 was collected from staff gauges read tri-daily (Coordinating Committee, 1978). Although the variability of this earlier data does not appear to be significantly different than the recording gauge data collected post-1950, the slope of the linear relationship in the earlier data does appear to be slightly different than the slope of the latter data. If the earlier data is omitted the downward slope of the regression line increases to -29.00 cm/century. 4.6. Two Harbors, Minnesota The water level record begins in 1887 at Two Harbors, but very little data is available until 1941. In addition, the water level record ends in 1988. The water level difference 11 plot for Point Iroquois minus Two Harbors is shown in Figure 4-4. The rate of vertical movement as determined from the slope of the linear regression line is -19.90 cm/century, indicating that Two Harbors is falling relative to the lake’s outlet. Visual inspection of the plot does not reveal any significant discontinuities. Point Iroquois - Two Harbors 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1920 1940 1960 1980 2000 Year Figure 4-4: Point Iroquois minus Two Harbors water level difference plot. 4.7. Grand Marais, Minnesota The water level record begins in 1966 at Grand Marais. The water level difference plot for Point Iroquois minus Grand Marais is shown in Figure 4-5. The rate of vertical movement as determined from the slope of the linear regression line is -10.26 cm/century, indicating that Grand Marais is falling relative to the lake’s outlet. Visual inspection of the plot does not reveal any significant discontinuities. 12 Point Iroquois - Grand Marais 0.3 Jun-Sep Avg Linear (Jun-Sep Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1980 2000 2020 Year Figure 4-5: Point Iroquois minus Grand Marais water level difference plot. 4.8. Thunder Bay, Ontario The water level record begins in 1907 at Thunder Bay. The water level difference plot for Point Iroquois minus Thunder Bay is shown in Figure 4-6. The rate of vertical movement as determined from the slope of the linear regression line is 3.84 cm/century, indicating that Thunder Bay is rising relative to the lake’s outlet. However, visual inspection of Figure 4-6 shows that though it is not entirely misleading, a linear relationship in this case may not be an accurate representation of the apparent vertical movement, since the plot shows some interesting discontinuities. Two alternative possibilities were investigated. The first was that the data prior to 1951 is inconsistent with the data recorded from 1951 on. The rate of vertical movement calculated with only data collected from 1951 on is 1.29 cm/century. The second possibility investigated was that a change in rate of vertical movement has occurred at some point during the period of record. A breakpoint analysis was performed, which identified a possible breakpoint in 1985 if the pre-1951 data is included, or 1989 if only data collected from 1951 onwards is used. However, there is little evidence suggesting reasons for either of these alternative possibilities. All data were collected from continuously recording gauges. A number of gauge relocations have occurred, but only one, at Thunder Bay in 1953, has occurred close to any of the dates of the discontinuities suggested by the additional analysis. This relocation was a short distance, and it seems unlikely that this was the cause of the 13 discontinuity. No other gauges on Lake Superior show similar issues with pre-1951 data, so it seems unlikely that the data at Point Iroquois is the problem. Comparisons of Thunder Bay versus other long-term gauges on Lake Superior (i.e. Duluth, Michipicoten, and Marquette) appear to show similar shifts in the pre-1951 data, though these are not as discernible due to the greater slope of the data in these plots. The mild slope of the Point Iroquois minus Thunder Bay plot makes the shift more noticeable. Lastly, the period of record after the possible breakpoints in the mid- to late-1980s is too short to adequately define a change in rate. Point Iroquois - Thunder Bay 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1951+) Water Level Difference (m) 0.2 0.1 0 -0.1 1920 1940 1960 1980 2000 2020 Year Figure 4-6: Point Iroquois minus Thunder Bay water level difference plot. 4.9. Rossport, Ontario The water level record begins in 1967 at Rossport. The water level difference plot for Point Iroquois minus Rossport is shown in Figure 4-7. The rate of vertical movement as determined from the slope of the linear regression line is 24.55 cm/century; however, visual inspection of the plot shows a significant jump occurring in 1973. Tushingham (1992) reports that the Rossport gauge has been tied to an unstable benchmark in the past. While this could not be confirmed, it seems a likely explanation for the observed jump, especially since the observed rate of apparent vertical movement is fairly constant after this point in time. The rate of apparent vertical movement as calculated from 1973 data onwards is 15.41 cm/century, which seems to be a more realistic estimate. 14 Point Iroquois - Rossport 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1973+) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 2020 Year Figure 4-7: Point Iroquois minus Rossport water level difference plot. 4.10. Michipicoten, Ontario The water level record begins in 1915 at Michipicoten. The water level difference plot for Point Iroquois minus Michipicoten is shown in Figure 4-8. The rate of vertical movement as determined from the slope of the linear regression line is 23.39 cm/century. Visual inspection of the plot does not reveal any significant discontinuities. A breakpoint analysis was performed, which suggested a potential change in rate around 1965. However, since it is not possible to confirm a rate change, and since the linear regression line for the entire period of record does provide what appears to be a reasonably good fit (R-squared = 0.968), the rate of 23.39 cm/century is acceptable. 15 Point Iroquois - Michipicoten 0.3 Jun-Sep Avg Linear (Jun-Sep A ) Water Level Difference (m) 0.2 0.1 0 -0.1 -0.2 1920 1940 1960 1980 2000 2020 Year Figure 4-8: Point Iroquois minus Michipicoten water level difference plot. 4.11. Gros Cap, Ontario The water level record begins in 1961 at Gros Cap. The water level difference plot for Point Iroquois minus Gros Cap is shown in Figure 4-9. The rate of vertical movement as determined from the slope of the linear regression line is -2.91 cm/century. However, visual inspection of the plot reveals what appear to be significant discontinuities or outliers in the period of record, especially in the early years. In more recent years (post1986), the water level difference relationship appears more stable. The rate of vertical movement calculated for this period is -1.11 cm/century. Though it represents a very short period of record, this rate is also more in line with what one would expect, since Gros Cap and Point Iroquois are located very close together. In fact, for many purposes, due to their close proximity, the rate calculated at Gros Cap with respect to Point Iroquois is likely redundant. 16 Point Iroquois - Gros Cap 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1986+) Water Level Difference (m) 0.2 0.1 0 -0.1 1940 1960 1980 2000 2020 Year Figure 4-9: Point Iroquois minus Gros Cap water level difference plot. 4.12. Summary The calculated rates of apparent vertical movement and their standard deviations for Lake Superior are given in Table 4-2. The water level gauge, period of record, and its calculated rate of apparent vertical movement and associated standard error with respect to the outlet of Lake Superior at Point Iroquois are given. Rates and standard error values calculated for additional scenarios are also given in italics. These scenarios were computed for the reasons outlined in the text above (i.e. where gauge issues were observed or suspected). Note that rates were also calculated with 1951 data onwards for all long-term gauges due to the observed issue with the pre-1951 data in the Thunder Bay minus Point Iroquois water level difference plot. Though this potential issue, if it exists, might be related to Thunder Bay only, the calculated rates with 1951 data onwards also provides an estimate of vertical movement with a consistent period of record between all gauges that does not include staff gauge data or any other potential pre-1951 data issues. In general, water level difference plots on Lake Superior are fairly uncomplicated. No single outliers were removed from the Lake Superior plots, although consecutive years of questionable data were removed for the analysis at both Rossport and Gros Cap. With the possible exception of Thunder Bay, water level difference plots on Lake Superior do not show any significant discontinuities or anomalies. Most of the water level difference plots on Lake Superior appear to be well-represented by a linear equation, 17 and the variability in the plots does not mask or lead one to question the observed trend in most cases. Gauge (Period of Record) Rate Standard Error *Additional Scenarios (cm/century) (cm/century) Marquette (1931-2006) -10.50 0.32 -9.78 0.45 *Marquette (1951-2006) Ontonagon (1960-2006) -17.32 1.27 Duluth (1931-2006) -26.85 0.53 -29.00 0.73 *Duluth (1951-2006) Two Harbors (1931-1987) -19.90 1.08 Grand Marais (1966-2006) -10.26 1.01 Thunder Bay (1931-2006) 3.84 0.70 1.29 1.06 *Thunder Bay (1951-2006) Rossport (1967-2006) 24.55 2.21 15.41 1.64 *Rossport (1973-2006) Michipicoten (1931-2006) 23.39 0.52 20.63 0.71 *Michipicoten (1951-2006) Gros Cap (1961-2006) -2.91 2.52 -1.11 1.33 *Gros Cap (1986-2006) Table 4-2: Summary of Lake Superior rates of apparent vertical movement and uncertainty relative to Point Iroquois. 5. Lakes Michigan-Huron 5.1. Overview Lake Michigan and Lake Huron are connected by the wide and deep Straits of Mackinac, which causes the two lakes to behave as one in terms of hydraulics. For this reason they are considered together in this analysis. Water level data from a total of eight gauge stations on Lake Michigan and twelve gauge stations on Lake Huron were analyzed, for a total of 20 gauges on the two lakes combined. On Lake Michigan all eight gauges are located within the United States, as are six of the gauges on Lake Huron. The remaining six gauge stations on Lake Huron are located within Canada. Table 5-1 and Table 5-2 give a list of the gauge stations used in this analysis, as well as a summary of the periods of record for each station, including any full years of data that were not available. 18 Figure 5-1: Lakes Michigan-Huron gauge locations. Station ID Location Start Year End Year Missing (Full) Years Ludington, MI, 1898-99; 1901; 1905; 1907; USA 1895 2006 1909-34; 1938; 1940-43; 1948-49 Holland, MI, 1898; 1901-02; 1904; 1909-34; 9087031 1894 2006 USA 1936-40; 1943-55; 1957-58 Calumet 9087044 -Harbor, IL, USA 1903 2006 Milwaukee, WI, 9087057/8 USA 1860 2006 -Kewaunee, WI, 9087068 1974 2006 -USA Sturgeon Bay 9087072 1920-21; 1923-24; 1926 Canal, WI, USA 1905 2006 Green Bay, WI, 9087078/9 USA 1953 2006 -Port Inland, MI, 9087096 1964 2006 -USA Table 5-1: Lake Michigan gauge stations and periods of record. 9087023 Number of Full Years Missing Total Years of Data 38 74 50 63 0 104 0 147 0 33 5 97 0 54 0 43 19 Station ID Location 9075002 9075014 9075034/5 9075059 9075080 Lakeport, MI, USA Harbor Beach, MI, USA Essexville, MI, USA Harrisville, MI, USA Mackinaw City, MI, USA 9075098/9 Start Year 1955 End Year 2006 1860 1953 1961 1899 Missing (Full) Years -- Number of Full Years Missing 0 Total Years of Data 52 2006 2006 2006 ---- 0 0 0 147 54 46 2006 -1897-98; 1900; 1904-33; 1937-43 0 108 40 71 0 81 0 48 0 47 2 99 0 45 0 97 De Tour, MI, USA 1896 2006 Thessalon, ON, Canada 11070 1926 2006 -Little Current, ON, 11195 1959 2006 -Canada Parry Sound, ON, 11375 1960 2006 -Canada Collingwood, ON, Canada 11500 1906 2006 1912-1913 Tobermory, ON, 11690 1962 2006 -Canada Goderich, ON, 11860 1910 2006 -Canada Table 5-2: Lake Huron gauge stations and periods of record. 5.2. Harbor Beach, Michigan The Harbor Beach gauge on Lake Huron has traditionally been selected as the outlet of the two lakes, and was originally selected in this analysis as well. The Lakeport gauge was also considered as it is located closer to the actual outlet at the head of the St. Clair River. The period of record for Lakeport is much shorter than that at Harbor Beach. For this reason Harbor Beach has traditionally been chosen as the outlet of Lakes MichiganHuron. However, the period of record at Lakeport now contains greater than 50 years of water level data, and, as will be shown in Section 5.11, there appears to be some movement between Lakeport and Harbor Beach. For these reasons and the close proximity of Lakeport to the head of the St. Clair River, Lakeport was also considered separately as the outlet of the two lakes in Section 5.23. A third gauge station, Goderich, is also located fairly close to the actual outlet, but potential gauge issues at this location as outlined below prevented it from being chosen. Data at Harbor Beach begins in 1860 and continues through to 2006; however, it has been noted by Quinn and Southam (2008) that water level data from 1860 to August 1874 is actually Milwaukee data transferred to Harbor Beach using an estimated rate of vertical movement of 14 cm/century. It should also be noted that data from September 1874 until March 1901 was collected from staff gauge readings, while data after this point in time 20 until present has been collected with a continually recording gauge. Both of these issues are discussed in greater detail below. 5.3. Ludington, Michigan The water level record begins in 1895 at Ludington; however, until 1950 monthly data is limited to only a handful of months, if any, having records in a given year. Furthermore, data from this period was collected using staff gauges. Only beginning in August 1950, after the first continuously recording gauge was installed in July of that year, does monthly data become continually available. The water level difference plot for Harbor Beach minus Ludington is shown in Figure 5-2. The rate of vertical movement as determined from the slope of the linear regression line is -12.77 cm/century. However, if one omits the pre-1951 data, the rate becomes -17.73 cm/century. For the reasons outlined above, it is suggested that this rate is the more representative estimate. Harbor Beach - Ludington 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1951+) Water Level Difference (m) 0.2 0.1 0 -0.1 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-2: Harbor Beach minus Ludington water level difference plot. Visual inspection of the plot followed by breakpoint analysis indicated a potential breakpoint at approximately 1971. The rate of movement calculated after 1971 is -13.81 cm/century. However, the period of record in this case is likely too short, and it 21 may simply be that the variability in the plot gives the impression of a breakpoint. Another alternative is that the breakpoint is actually a sudden drop. Evidence of a drop is also apparent when Ludington is plotted versus other long-term Lakes Michigan-Huron gauges, including Milwaukee (Figure 5-3), Mackinaw City and Calumet Harbor. However, no evidence has been found to support a possible reason for either a change in rate or a sudden jump at the Ludington station. Milwaukee - Ludington 0.3 Jun-Sep Avg Water Level Difference (m) 0.2 0.1 0 -0.1 1940 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-3: Milwaukee minus Ludington water level difference plot. 5.4. Holland, Michigan The water level record begins in 1894 at Holland; however, similar to Ludington, monthly data is limited in the early period, and was collected using staff gauges. Only beginning in May 1959, after the first continuously recording gauge was installed at this location, does monthly data become continually available. The water level difference plot for Harbor Beach minus Holland is shown in Figure 5-4. The rate of vertical movement as determined from the slope of the linear regression line is -9.16 cm/century. If one omits the pre-1959 data, however, the rate becomes -11.70 cm/century. For the reasons outlined above, it is suggested that this rate is the more representative estimate. Visual inspection does not show any significant discontinuities. 22 Harbor Beach - Holland 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1959+) Water Level Difference (m) 0.2 0.1 0 -0.1 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-4: Harbor Beach minus Holland water level difference plot. 5.5. Calumet Harbor, Illinois The water level record begins in 1903 at Calumet Harbor. The water level difference plot for Harbor Beach minus Calumet Harbor was shown previously in Figure 3-2. A visual inspection of this plot shows what appears to be an outlier occurring in 1907. The June to September monthly average difference in 1907 appears to disagree significantly with the other data, and was therefore not used in this analysis. The rate of vertical movement as determined from the slope of the linear regression line with the 1907 outlier removed (Figure 5-5) is -12.12 cm/century, which is only slightly smaller than the rate calculated with the outlier included (-12.92 cm/century). For the reasons outlined above, it is suggested that the rate of -12.12 cm/century is the more reasonable estimate. A visual inspection of the data and an additional plot of the five-year centred running average (Figure 5-6) also indicate what appears to be a stepwise pattern to the data. As will be shown below, this pattern is similar to other patterns observed on Lake Michigan, particularly at Milwaukee. 23 Harbor Beach - Calumet Harbor 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-5: Harbor Beach minus Calumet Harbor water level difference plot with 1907 outlier removed. Harbor Beach - Calumet Harbor 0.3 June-Sep. Avg. June-Sep. 5-Year Centred Avg. Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-6: Harbor Beach minus Calumet Harbor water level difference plot with 5-year centred running average. 24 5.6. Milwaukee, Wisconsin The water level record begins in 1860 at Milwaukee. The water level difference plot for Harbor Beach minus Milwaukee is shown in Figure 5-7. Quinn and Southam (2008) have outlined a number of issues with this plot, which can also be seen here. First, as was discussed in Section 5.2, the water level data at Harbor Beach for the period from 1960 to August 1874 is actually Milwaukee data transferred to Harbor Beach using an estimated rate of vertical movement of 14 cm/century (i.e. the Harbor Beach data for this period was not measured at Harbor Beach). This can be seen clearly by the reduced variability during this period in Figure 5-7. Furthermore, data collected at Milwaukee after this period until September 1903 was collected from staff gauge readings. From this point on, except for a short period from March 1946 until July 1951, recording gauges have been in use. The water level differences appear to show greater variability during the staff gauge period than during the recording gauge period, at least prior to 1903. This increase in variability is not as evident in the shorter staff gauge period that began in March 1946. Harbor Beach - Milwaukee 0.3 Jun-Sep Avg 1875 Linear (1904+) 1904 Water Level Difference (m) 0.2 Linear (1875+) 0.1 0 -0.1 1840 1860 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-7: Harbor Beach minus Milwaukee water level difference plot. The early data (up until 1874) transferred from Milwaukee was discarded from this analysis, and the average rate of apparent vertical movement was calculated for both 1875 data onwards and 1904 data onwards. The rate calculated from 1875 on, which includes the staff gauge data, is -14.77 cm/century. The rate calculated from 1904 on, which does not include the staff gauge data, is -16.14 cm/century. 25 A plot of the five-year centred running average for Harbor Beach minus Milwaukee was previously shown in Figure 3-3. It indicates what appears to be a stepwise pattern to the water level differences. Interestingly, the stepwise pattern observed at Milwaukee is similar to the patterns seen at other gauges on Lake Michigan (Figure 5-8). The pattern at Calumet Harbor in particular appears to be similar to that observed at Milwaukee. In addition, Sturgeon Bay Canal and Mackinaw City, though not quite as apparent, also appear to have steps at similar places, such as the one seen around 1960. Harbor Beach - Lake Michigan Gauges 5-year Centred Running Average 0.3 HB-MIL HB-CAL HB-StB Water Level Difference (m) 0.2 HB-MAC 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-8: Harbor Beach minus various Lake Michigan gauges 5-year centred running averages of water level difference plots: HB = Harbor Beach; MIL = Milwaukee; CAL = Calumet Harbor; StB = Sturgeon Bay Canal; MAC = Mackinaw City. One possible explanation is that the Harbor Beach data is the cause of the pattern observed; however, the pattern is not seen at gauges on Lake Huron when compared to Harbor Beach, as would be expected if this was the case (although this might be the result of some other issue on Lake Huron, which is discussed below). Another explanation for the similar patterns is that the gauges, in particular Milwaukee and Calumet Harbor, are susceptible to the same forces, whether gauge-related, meteorological or geological in nature. For example, perhaps both gauges tend to be affected by the same storm events, or both locations are actually falling with respect to Harbor Beach in a stepwise pattern, as opposed to linearly, due to glacial isostatic adjustment. The proximity of the two gauges lends credence to these possible explanations. While a detailed investigation into the root causes of the stepwise effects is beyond the scope of this document, the pattern highlighted should be taken into consideration in future studies, including measurement 26 programs involving Global Positioning Systems (GPS) where short-term movement rates are important. 5.7. Kewaunee, Wisconsin The water level record begins in 1974 at Kewaunee. With only 33 years of data, this is a relatively short period of record. The water level difference plot for Harbor Beach minus Kewaunee is shown in Figure 5-9. The rate of vertical movement as determined from the slope of the linear regression line is -11.99 cm/century; however, rates calculated from two other four month averages (i.e. May to August and July to October) showed significantly different results (-13.65 and -9.75 cm/century, respectively). The regression line in Figure 5-9 is heavily influenced by the higher data occurring before 1977, and the apparent outlier occurring in 2002. With this data excluded the rate becomes only -4.19 cm/century; however, there is no evidence to suggest any reason for the exclusion of this data, especially given the short period of record. The short period of record does not appear to be sufficient for estimating the rate of vertical movement in this case, and in future years, as more data is added to the Kewaunee period of record, a better understanding of the vertical movement occurring at this gauge may become apparent. Harbor Beach - Kewaunee 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1970 1975 1980 1985 1990 1995 2000 2005 2010 Year Figure 5-9: Harbor Beach minus Kewaunee water level difference plot. 27 5.8. Sturgeon Bay Canal, Wisconsin The water level record begins in 1905 at Sturgeon Bay Canal, but there is a gap in the record from April 1919 to June 1927, and the gauge location was moved during this time. Furthermore, data collected before June 1945 is staff gauge data, while data from this month onwards is collected from a continuously recording gauge. The water level difference plot for Harbor Beach minus Sturgeon Bay Canal is shown in Figure 5-10. The rate of vertical movement as determined from the slope of the linear regression line is -4.36 cm/century, indicating Sturgeon Bay Canal is falling relative to Harbor Beach. Harbor Beach - Sturgeon Bay Canal 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1927+) Water Level Difference (m) 0.2 Linear (1945+) 0.1 0 -0.1 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-10: Harbor Beach minus Sturgeon Bay Canal water level difference plot. Visual inspection of the plot, however, shows at least two additional features. First, the early data collected from 1905 until the missing period that begins in 1919 does not appear to be well represented by a linear regression line fitted over the entire period of record. The rate of vertical movement with this early data excluded is -6.73 cm/century. Second, whether this early data is included or not, there appears to be a breakpoint, which according to the breakpoint analysis occurs in the mid to late 1940s. Since the data collection method was changed to a recording gauge in 1945, this year was selected as the breakpoint. The rate of vertical movement calculated with recording gauge data only 28 (1945 onwards) is -8.26 cm/century. This regression line most closely represents the apparent vertical movement occurring at this location in recent time. 5.9. Green Bay, Wisconsin The water level record begins in 1953 at Green Bay. The water level difference plot for Harbor Beach minus Green Bay is shown in Figure 5-11. The rate of vertical movement as determined from the slope of the linear regression line is -9.37 cm/century, indicating Green Bay is falling relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. Harbor Beach - Green Bay 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-11: Harbor Beach minus Green Bay water level difference plot. 5.10. Port Inland, Michigan The water level record begins in 1964 at Port Inland. The water level difference plot for Harbor Beach minus Port Inland is shown in Figure 5-12. The rate of vertical movement as determined from the slope of the linear regression line is 6.03 cm/century, indicating Green Bay is rising relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. 29 Harbor Beach - Port Inland 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 5-12: Harbor Beach minus Port Inland water level difference plot. 5.11. Lakeport, Michigan The water level record begins in 1955 at Lakeport. As stated in Section 5.2, Lakeport is located closer to the outlet of Lakes Michigan-Huron at the head of the St. Clair River than Harbor Beach is. Harbor Beach has traditionally been chosen as the outlet of the two lakes due to its location and longer period of record. However, since Lakeport now has greater than 50 years of data, it was also considered as the outlet in this analysis (Section 5.23). As more data is added to the period of record in coming years, Lakeport may become the preferred choice to represent the Lakes Michigan-Huron outlet. The water level difference plot for Harbor Beach minus Lakeport is shown in Figure 5-13. The data for 1976 was removed as an outlier, as there was an obvious problem with the water level gauge at Lakeport during June and July of this year, which resulted in unrealistic recorded water levels that were between 0.6 and 0.7 metres less than those recorded in the months of May and August. With this year removed, the rate of vertical movement as determined from the slope of the linear regression line is -3.19 cm/century, indicating that Lakeport is falling relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. While the rate calculated is not large, it does indicate that Lakeport is falling relative to Harbor Beach, and that Harbor Beach may not be as good a representation of the outlet 30 as has been previously believed. Further discussion on Lakeport as the outlet is given in Section 5.23. Harbor Beach - Lakeport 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-13: Harbor Beach minus Lakeport water level difference plot. 5.12. Essexville, Michigan The water level record begins in 1953 at Essexville. The water level difference plot for Harbor Beach minus Essexville is shown in Figure 5-14. The rate of vertical movement as determined from the slope of the linear regression line is -5.58 cm/century, indicating that Essexville is falling relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. 31 Harbor Beach - Essexville 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-14: Harbor Beach minus Essexville water level difference plot. 5.13. Harrisville, Michigan The water level record begins in 1961 at Harrisville. The water level difference plot for Harbor Beach minus Harrisville is shown in Figure 5-15. The rate of vertical movement as determined from the slope of the linear regression line is 5.48 cm/century, indicating that Harrisville is rising relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. In fact, Harbor Beach minus Harrisville provides an excellent example of how gauges located in close proximity can show little variance in terms of water level differences, as the plot shows that the linear regression line appears to fit the data well. 32 Harbor Beach - Harrisville 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 5-15: Harbor Beach minus Harrisville water level difference plot. 5.14. Mackinaw City, Michigan The water level record begins in 1899 at Mackinaw City. The water level difference plot for Harbor Beach minus Mackinaw City is shown in Figure 5-16. The rate of vertical movement as determined from the slope of the linear regression line is 9.22 cm/century, indicating that Mackinaw City is rising relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. A plot of the 5-year centred running average does appear to indicate a stepped pattern (Figure 5-17), though not a significant one, and this observation may only be the result of variability in the water level difference relationship. 33 Harbor Beach - Mackinaw City 0.1 Water Level Difference (m) 0 -0.1 -0.2 Jun-Sep Avg Linear (Jun-Sep Avg) -0.3 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-16: Harbor Beach minus Mackinaw City water level difference plot. Harbor Beach - Mackinaw City 0.1 Water Level Difference (m) 0 -0.1 -0.2 June-Sep. Avg. June-Sep. 5-Year Centred Avg. -0.3 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-17: Harbor Beach minus Mackinaw City water level difference plot with 5-year centred running average. 34 5.15. De Tour, Michigan The water level record begins in 1896 at De Tour; however, monthly data is limited to only a handful of months, if any, having records in a given year until 1951. Only beginning in 1951 does monthly data become continually available. Furthermore, data was collected using staff gauges until June 1954. The water level difference plot for Harbor Beach minus De Tour is shown in Figure 5-18. The rate of vertical movement as determined from the slope of the linear regression line is 14.63 cm/century. In this case omitting the pre-1954 staff gauge data made little difference to the calculated rate (14.15 cm/century with recording gauge data only). There do not appear to be any significant discontinuities in the plot. Harbor Beach - De Tour 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-18: Harbor Beach minus De Tour water level difference plot. 5.16. Thessalon, Ontario The water level record begins in 1926 at Thessalon. The water level difference plot for Harbor Beach minus Thessalon is shown in Figure 5-19. The rate of vertical movement as determined from the slope of the linear regression line is 18.60 cm/century, indicating that Thessalon is rising relative to Harbor Beach. 35 Harbor Beach - Thessalon 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1961-) Water Level Difference (m) 0.2 Linear (1980+) 0.1 0 -0.1 -0.2 1920 1940 1960 1980 2000 2020 Year Figure 5-19: Harbor Beach minus Thessalon water level difference plot. Visual inspection of the plot does not show any significant discontinuities; however, a plot of the 5-year centred running average does appear to indicate a stepped pattern, or plateau, occurring approximately during the 1960s and 1970s (Figure 5-20). Of note is that if two regression lines are drawn both before and after the apparent plateau, the rate of apparent vertical movement is greater than that calculated for the full data set, and approximately equal (21.73 and 20.85 cm/century, respectively), suggesting two approximately parallel periods of movement. However, it must also be noted that the variability in the plot appears to be greatest during the plateau period, and that this may in fact be the cause for any semblance of a plateau in the first place. This period should be investigated further. 36 Harbor Beach - Thessalon 0.1 Water Level Difference (m) 0 -0.1 -0.2 June-Sep. Avg. June-Sep. 5-Year Centred Avg. -0.3 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Year Figure 5-20: Harbor Beach minus Thessalon water level difference plot with 5-year centred running average. 5.17. Little Current, Ontario The water level record begins in 1959 at Little Current. The water level difference plot for Harbor Beach minus Little Current is shown in Figure 5-21. The rate of vertical movement as determined from the slope of the linear regression line is 25.93 cm/century. However, visual inspection of the plot shows a significant jump occurring in 1963. Tushingham (1992) reports that the Little Current gauge has been tied to an unstable benchmark in the past. The same issue was reported by Tushingham for the Rossport gauge on Lake Superior (Section 4.9), and the plots of both Rossport and Little Current show a similar looking jump. While it could not be confirmed, it seems likely that the unstable benchmark explains the observed jump at both these gauges. Like Rossport, the observed rate of apparent vertical movement is fairly constant at Little Current after this point in time. The rate as calculated from post-1963 data is 21.33 cm/century, which is likely a more realistic estimate of the apparent vertical movement at this location. 37 Harbor Beach - Little Current 0.1 Water Level Difference (m) 0 -0.1 -0.2 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1963+) -0.3 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-21: Harbor Beach minus Little Current water level difference plot. 5.18. Parry Sound, Ontario The water level record begins in 1960 at Parry Sound. The water level difference plot for Harbor Beach minus Parry Sound is shown in Figure 5-22. The rate of vertical movement as determined from the slope of the linear regression line is 20.80 cm/century, indicating that Parry Sound is rising relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities. 38 Harbor Beach - Parry Sound 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-22: Harbor Beach minus Parry Sound water level difference plot. 5.19. Collingwood, Ontario The water level record begins in 1906 at Collingwood. The water level difference plot for Harbor Beach minus Collingwood is shown in Figure 5-23. The rate of vertical movement as determined from the slope of the linear regression line is 14.21 cm/century, indicating that Collingwood is rising relative to Harbor Beach. However, visual inspection of the plot shows some interesting features. These are especially evident in the 5-year centred running average plot (Figure 5-24). First, the pre1916 data does not appear to fit the rest of the data plot. This may be the result of the gauge being moved in 1915 (Coordinating Committee, 1978). Second, the assumption of a linear trend in this case appears to be invalid. Rather, the slope of the plot appears to shift at some point, or is possibly even curvilinear. Breakpoint analysis estimated a breakpoint to occur in 1967 (Figure 5-25). The slope calculated prior to the breakpoint is 19.14 cm/century, while the slope calculated after the breakpoint is 7.65 cm/century. One possible explanation is that the water level gauge at Collingwood was moved in 1965 (Coordinating Committee, 1978), although this is just one possibility, and since similar issues are seen at Goderich (Section 5.21), additional forces may be the cause. 39 Harbor Beach - Collingwood 0.1 Water Level Difference (m) 0 -0.1 1900 1920 1940 1960 1980 2000 2020 -0.2 Jun-Sep Avg Linear (Jun-Sep Avg) -0.3 Year Figure 5-23: Harbor Beach minus Collingwood water level difference plot. Harbor Beach - Collingwood 0.1 Water Level Difference (m) 0 -0.1 -0.2 June-Sep. Avg. June-Sep. 5-Year Centred Avg. -0.3 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-24: Harbor Beach minus Collingwood water level difference plot with 5-year running average. 40 Harbor Beach - Collingwood 0.1 Water Level Difference (m) 0 1967 -0.1 -0.2 June to Sep Avg. Piecewise Regression Estimated BreakPoint -0.3 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-25: Harbor Beach minus Collingwood breakpoint estimation. 5.20. Tobermory, Ontario The water level record begins in 1962 at Tobermory. The water level difference plot for Harbor Beach minus Tobermory is shown in Figure 5-26. The rate of vertical movement as determined from the slope of the linear regression line is 15.74 cm/century, indicating that Tobermory is rising relative to Harbor Beach. Visual inspection of the plot does not show any significant discontinuities, with the possible exception of a breakpoint occurring around 1969. This point coincides approximately with the breakpoint observed at Collingwood, although the direction is reversed. It could simply be the result of variability in the water level difference plot, however, and due to the short period of record it would not be prudent to make any more significant statements in this regard. 41 Harbor Beach - Tobermory 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 5-26: Harbor Beach minus Tobermory water level difference plot. 5.21. Goderich, Ontario The water level record begins in 1910 at Goderich; however, water level data appears only to have been collected during the summer and fall months until 1920, although twice during this period monthly records were kept until December. Furthermore, the gauge at Goderich was relocated in 1919, and has been moved a number of additional times since then (Coordinating Committee, 1978). The water level difference plot for Harbor Beach minus Goderich is shown in Figure 5-27. The rate of vertical movement as determined from the slope of the linear regression line is -2.97 cm/century, indicating that Goderich is falling relative to Harbor Beach. However, visual inspection of the plot shows some significant features. First, the pre1920 data does not appear to fit the rest of the data plot very well, which may be the result of the gauge relocation in 1919 (although, as noted, there have been several other gauge relocations at this site that do not show similar effects). 42 Harbor Beach - Goderich 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0.0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-27: Harbor Beach minus Goderich water level difference plot. Second, from approximately 1966 to 1978, the variability in the water level difference relationship appears to be significantly greater than the variability throughout the remainder of the plot. Interestingly, this period of increased variability seems to coincide with a similar period of interest in the Harbor Beach minus Thessalon water level difference relationship, during which there appeared to be either a pause in movement or the impression of a pause due to increased variability in the plot (Section 5.16). A plot of Goderich minus Thessalon is shown in Figure 5-28. This plot shows the same increase in variability occurring approximately during the 1970s as was seen in the plots of both Goderich and Thessalon versus Harbor Beach. Also, while the regression line in Figure 5-28 appears to be a relatively good fit, breakpoint analysis indicated that a change in slope of approximately 6 cm/century may have occurred sometime during the 1960s. The breakpoint, however, is not as obvious in this case due to the greater magnitude of the slope values calculated before and after the point in question. In fact, other plots can show similar traits, in which breakpoint analysis estimates a point indicating what would seem to be a significant change in slope numerically, but visual inspection of the plot does not readily reveal the same estimated breakpoint. This is a result of the variability in these plots, and it shows that while both visual inspection and the breakpoint analysis can be useful in indicating a potential breakpoint, the decision as to whether a break has actually occurred is somewhat subjective. 43 Goderich - Thessalon 0.1 Water Level Difference (m) 0 -0.1 -0.2 Jun-Sep Avg Linear (Jun-Sep Avg) -0.3 1920 1940 1960 1980 2000 2020 Year Figure 5-28: Goderich minus Thessalon water level difference plot. Finally, similar to observations made at the Collingwood gauge, the assumption of a linear trend in the case of Goderich appears to be invalid. Rather, the slope of the plot appears to shift at some point, or is possibly even curvilinear. As shown earlier in Section 3.5, breakpoint analysis estimated a breakpoint to occur in 1955. Considering the variability in the plots and the inexact nature of identifying the breakpoint, this value for Goderich is not far removed from the breakpoint of 1967 identified at Collingwood (Figure 5-25). The slope calculated in the Harbor Beach minus Goderich plot prior to the breakpoint is 1.22 cm/century, while the slope calculated after the breakpoint is -6.51 cm/century, indicating that Goderich went from rising relative to Harbor Beach, to falling relative to the lake’s outlet. One possible explanation is that salt mining, which began in the late 1950s at Goderich, may be causing the land in the area to subside. However, this explanation would apply to Goderich only, and since similar issues are seen at Collingwood, this may indicate that additional forces are the cause. A plot of Goderich minus Collingwood water levels tends to support this contention (Figure 5-29). The slope of the linear regression line is 17.21 cm/century, and as can be seen from the plot, a breakpoint is not evident. In fact, the slopes of linear regression lines calculated before and after a breakpoint of 1955 are within 1 cm/century of each other. This implies that not only do both Collingwood and Goderich show a breakpoint occurring at approximately the same time in the linear regression plot when compared to Harbor Beach, but also the change in 44 the slope value after the breakpoint is approximately equal (i.e. the slope of the plot decreases by approximately the same amount in both cases). A plot of Goderich minus Sturgeon Bay Canal (not shown) showed a similar result. Goderich - Collingwood 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 -0.2 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-29: Goderich minus Collingwood water level difference plot. Unfortunately the periods of records at other nearby gauges, such as Tobermory and Parry Sound for example, are not long enough to lend any further evidence to support a shift in the rate of apparent vertical movement. Nevertheless, as one last example to further illustrate the complexity inherent in this analysis, Figure 5-30 shows a plot of Milwaukee minus Goderich water levels. As can be seen, there does not appear to be a noticeable breakpoint, and in fact breakpoint analysis did not indicate any significant changes in slope. A stepwise pattern, however, similar to those seen in previous Milwaukee plots, is somewhat evident, as is larger year-to-year variability. 45 Goderich - Milwaukee 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 5-30: Goderich minus Milwaukee water level difference plot. 5.22. Summary Water level difference plots between gauge pairs on Lakes Michigan-Huron tended to be more complex than those plots observed on Lake Superior. While water level differences at some gauge pairs follow a mostly linear relationship, others, including Milwaukee and Calumet Harbor, appear to follow a more stepwise pattern. Others, such as Goderich and Collingwood, appear to show a shift in the rate of apparent vertical movement. Possible explanations include climate factors, mining activities, and actual geologic forces among others. Yet even when an explanation seems plausible (e.g. salt mining at Goderich as the cause of a change in rate of apparent vertical movement between Harbor Beach and Goderich), evidence from another gauge pair can be contradictory (e.g. Goderich minus Thessalon and Goderich minus Collingwood do not show the same change). While a full investigation of these issues is beyond the scope of this document, what this demonstrates is that our understanding of the underlying forces behind these observations is still incomplete. A summary of the calculated rates of apparent vertical movement for Lakes Michigan-Huron relative to Harbor Beach are shown in Table 5-3. The uncertainty (standard error of the slope value) is also shown, as are additional scenarios as described in previous sections. 46 Gauge (Period of Record) Rate Standard Error *Additional Scenarios (cm/century) (cm/century) Ludington (1897-2006) -12.77 0.77 -17.73 0.83 *Ludington (1951-2006) -13.81 1.55 *Ludington (1971-2006) Holland (1906-2006) -9.16 0.58 -11.70 0.92 *Holland (1959-2006) Calumet Harbor (1903-2006) -12.92 0.81 -12.12 0.51 *Calumet Harbor (w/o 1907 outlier) Milwaukee (1860-2006) -14.33 0.30 -14.77 0.36 *Milwaukee (1875-2006) -16.14 0.41 *Milwaukee (1904-2006) Kewaunee (1974-2006) -11.99 2.82 -4.19 1.54 *Kewaunee (1977-2006 & w/o 2002 outlier) Sturgeon Bay Canal (1905-2006) -4.36 0.44 -6.73 0.48 *Sturgeon Bay Canal (1927-2006) -8.26 0.57 *Sturgeon Bay Canal (1945-2006) Green Bay (1955-2006) -9.37 0.93 Port Inland (1965-2006) 6.03 1.21 Lakeport (1956-2006) -3.19 0.67 Essexville (1953-2006) -5.58 0.71 Harrisville (1962-2006) 5.48 0.52 Mackinaw City (1899-2006) 9.22 0.32 De Tour (1934-2006) 14.63 0.53 Thessalon (1927-2006) 18.60 0.41 21.73 0.90 *Thessalon (1927-1961) 20.85 1.60 *Thessalon (1980-2006) Little Current (1960-2006) 25.93 1.73 21.33 1.03 *Little Current (1963-2006) Parry Sound (1960-2006) 20.80 0.94 Collingwood (1906-2006) 14.21 0.36 19.14 0.63 *Collingwood (1916-1967) 7.65 1.10 *Collingwood (1967-2006) Tobermory (1962-2006) 15.74 1.08 Goderich (1910-2006) -2.97 0.41 1.22 0.88 *Goderich (1910-1955) -6.51 1.10 *Goderich (1955-2006) Table 5-3: Summary of Lakes Michigan-Huron rates of apparent vertical movement and uncertainty relative to Harbor Beach. 5.23. Outlet at Lakeport As discussed in Sections 5.2 and 5.11, Lakeport may be a better choice of outlet for Lakes Michigan-Huron than Harbor Beach, which has traditionally been chosen as the outlet. Lakeport is located closer to the true outlet at the head of the St. Clair River and 47 according to the water level difference plots Lakeport appears to be falling relative to Harbor Beach at a rate of -3.19 cm/century. In addition, water levels have now been recorded at Lakeport for more than 50 years, which is likely a sufficient period of record for calculating rates of apparent vertical movement, at least for the most recent time period. For these reasons, the rates of apparent vertical movement were also calculated with Lakeport chosen as the outlet of Lakes Michigan-Huron. Relative rates of movement with respect to both Harbor Beach and Lakeport are shown in Table 5-4 for comparison purposes. The rates shown for Harbor Beach were calculated for the same period of record as those for Lakeport. The period of record begins in 1955 at Lakeport, but the June to September averages only begin in 1956. Rate of Movement Difference (cm/century) Net Sum Gauge (Period of Record) in Rate (Diff. – (HB-LP rate)) Harbor (cm/cent.) Lakeport Beach -16.30 -13.11 -3.19 0.00 Ludington (1956-2006) -11.25 -7.54 -3.71 -0.52 Holland (1956-2006) -12.91 -9.86 -3.05 0.14 Calumet Harbor (1956-2006) -17.92 -14.59 -3.33 -0.14 Milwaukee (1956-2006) -11.99 -6.55 -5.44 -2.25 Kewaunee (1974-2006) -7.99 -4.37 -3.62 -0.43 Sturgeon Bay Canal (1956-2006) -9.68 -6.24 -3.44 -0.25 Green Bay (1956-2006) 6.03 9.99 -3.96 -0.77 Port Inland (1965-2006) -5.78 -2.55 -3.23 -0.04 Essexville (1956-2006) 5.48 8.20 -2.72 0.47 Harrisville (1962-2006) 10.21 13.72 -3.51 -0.32 Mackinaw City (1956-2006) 14.57 17.81 -3.24 -0.05 De Tour (1956-2006) 19.14 22.50 -3.36 -0.17 Thessalon (1956-2006) 25.93 28.91 -2.98 0.21 Little Current (1960-2006) 20.80 23.57 -2.77 0.42 Parry Sound (1960-2006) 9.04 12.11 -3.07 0.12 Collingwood (1956-2006) 15.74 18.20 -2.46 0.73 Tobermory (1962-2006) -6.27 -2.77 -3.50 -0.31 Goderich (1956-2006) Average -3.37 -0.18 Table 5-4: Comparison of rates of movement on Lakes Michigan-Huron relative to Harbor Beach and Lakeport. The key point to note from this table is the differences between Harbor Beach and Lakeport and the net sum of the differences. Note that if one triangulates between any three gauges on a given lake, the net sum of the differences moving around the triangle should be zero. For example, the difference between the rates of movement between Ludington and Harbor Beach and Ludington and Lakeport should theoretically be equal to the rate of movement between Harbor Beach and Lakeport, which was calculated as -3.19 cm/century in Section 5.11. In the case of Ludington the net sum is in fact zero. However, due to uncertainty in the calculated rates of apparent vertical movement, the 48 net sum does not end up being exactly zero at all gauges. Nevertheless, the differences between the rates calculated relative to Harbor Beach and those calculated relative to Lakeport are all fairly close to each other, and the average difference in rates is equal to -3.37 cm/century, which is quite close to the rate of movement of -3.19 cm/century calculated between Harbor Beach and Lakeport directly. These results provide further evidence that Lakeport is falling relative to Harbor Beach at a rate of approximately -3 cm/century or so. Again, since the period of record at Lakeport is now greater than 50 years, and since Lakeport is located much closer to the actual lake outlet, it should be given consideration for use as the outlet in future studies. Many of the issues and scenarios described in previous sections for Harbor Beach apply to Lakeport as well, so a full treatment of these is not given again here, but the rates of movement calculated for Lakeport and their uncertainty is provided in Table 5-5 for each gauge and a selection of similar scenarios to those outlined for Harbor Beach. Note that some scenarios investigated for Harbor Beach are not included since they occur prior to the start of the Lakeport period of record, or only shortly thereafter. Standard Error (cm/century) Ludington (1956-2006) -13.11 1.10 *Ludington (1971-2006) -7.81 1.39 Holland (1956-2006) -7.54 1.07 *Holland (1959-2006) -8.07 1.08 Calumet Harbor (1956-2006) -9.86 1.13 Milwaukee (1956-2006) -14.59 1.05 Kewaunee (1974-2006) -6.55 2.81 *Kewaunee (1977-2006 & w/o 2002 outlier) -1.16 1.95 Sturgeon Bay Canal (1956-2006) -4.37 1.01 Green Bay (1956-2006) -6.24 1.09 Port Inland (1965-2006) 9.99 1.40 Essexville (1956-2006) -2.55 0.85 Harrisville (1962-2006) 8.20 1.12 Mackinaw City (1956-2006) 13.72 0.99 De Tour (1956-2006) 17.81 1.08 Thessalon (1956-2006) 22.50 1.29 *Thessalon (1980-2006) 24.46 2.57 Little Current (1960-2006) 28.91 2.02 *Little Current (1963-2006) 24.27 1.50 Parry Sound (1960-2006) 23.57 1.29 Collingwood (1956-2006) 12.11 1.02 *Collingwood (1967-2006) 12.18 1.45 Tobermory (1962-2006) 18.20 1.60 Goderich (1956-2006) -2.77 1.32 Table 5-5: Summary of Lakes Michigan-Huron rates of apparent vertical movement and uncertainty relative to Lakeport. Gauge (Period of Record) *Additional Scenarios Rate (cm/century) 49 6. Lake Erie 6.1. Overview Water level data from a total of 16 gauge stations were analyzed on Lake Erie (Figure 6-1). Six of these gauge stations are located in Canada, while the remaining ten gauge stations are located in the United States. Table 6-1 gives a list of the gauge stations used in this analysis, as well as a summary of the periods of record for each station. Figure 6-1: Lake Erie gauge locations. Station ID 9063020 9063028 9063032 9063038 9063053 9063063 9063079 9063085 9063087 Location Buffalo, NY, USA Sturgeon Point, NY, USA Barcelona, NY, USA Erie, PA, USA Fairport, OH, USA Cleveland, OH, USA Marblehead, OH, USA Start Year 1860 1869 1960 1958 1975 1860 1959 End Year 2006 2006 1987 2006 2006 2006 2006 Missing (Full) Years 1870-1886 ------1878-1903; 1909-10 -- Toledo, OH, USA 1877 2006 Monroe, MI, USA 1975 1988 Fermi Power Plant, MI, 9063090 USA 1963 2006 -12005 Bar Point, ON, Canada 1966 2006 -12065 Kingsville, ON, Canada 1962 2006 -12250 Erieau, ON, Canada 1957 2006 -12400 Port Stanley, ON, Canada 1908 2006 1912-1925 12710 Port Dover, ON, Canada 1958 2006 -Port Colborne, ON, 12865 1911 2006 -Canada Table 6-1: Lake Erie gauge stations and periods of record. Number of Full Years Missing 17 0 0 0 0 0 0 Total Years of Data 130 138 28 49 32 147 48 28 0 102 14 0 0 0 0 14 0 44 41 45 50 85 49 0 96 50 6.2. Buffalo, New York The Buffalo gauge was selected as the outlet of Lake Erie. The Buffalo gauge is located close to the head of the Niagara River, has traditionally been used as the outlet and has a relatively long period of record. Buffalo data begins in 1860, but early data is incomplete and only available during the months of April to December until 1870, and from 1871 until 1887 there is a gap in which no data is available. Data only becomes continuously available from 1887 onwards. Moreover, data up until 1899 is staff gauge data, while continuously recording gauges were only used from this point onwards. 6.3. Sturgeon Point, New York The water level record begins in 1969 at Sturgeon Point. With only 38 years of data, this is a relatively short period of record. The water level difference plot for Buffalo minus Sturgeon Point is shown in Figure 6-2. The rate of vertical movement as determined from the slope of the linear regression line is 1.14 cm/century, indicating that Sturgeon Point is rising slightly relative to Buffalo. However, the short period of record certainly affects the accuracy of the estimate, and in future years, as more data is added to the Sturgeon Point period of record, a better understanding of the vertical movement occurring at this gauge may become apparent. Buffalo - Sturgeon Point 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 6-2: Buffalo minus Sturgeon Point water level difference plot. 51 6.4. Barcelona, New York The water level record begins in 1960 at Barcelona, but ends in 1987. As was the case at Sturgeon Point, with only 28 years of data, this is a relatively short period of record. The water level difference plot for Buffalo minus Barcelona is shown in Figure 6-3. The rate of vertical movement as determined from the slope of the linear regression line is -6.43 cm/century, indicating that Barcelona is falling relative to Buffalo. However, rates calculated for other combinations of months were significantly different (for example, the rate calculated taking the annual average of all monthly data was -3.49 cm/century). The short period of record certainly affects the accuracy of the estimate, and unfortunately, the Barcelona gauge has been discontinued, so it is unlikely that additional water level data will be added to the period of record. Since the calculated rate of apparent vertical movement appears questionable for these reasons, the inclusion of Barcelona should be questioned, if not omitted, in any additional studies on glacial isostatic adjustment. Buffalo - Barcelona 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 Year Figure 6-3: Buffalo minus Barcelona water level difference plot. 6.5. Erie, Pennsylvania The available water level record begins in 1958 at Erie, Pennsylvania. The water level difference plot for Buffalo minus Erie is shown in Figure 6-4. The rate of vertical 52 movement as determined from the slope of the linear regression line is -7.69 cm/century, indicating that Erie is falling relative to Buffalo. Visual inspection of the plot shows a break in the available data in 1984, which also gives an illusion of a possible step or drop at this point. A plot of the July to October averages, however, does not show evidence of this same issue. Buffalo - Erie, PA 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-4: Buffalo minus Erie, PA, water level difference plot. 6.6. Fairport, Ohio The available water level record begins in 1975 at Fairport. The water level difference plot for Buffalo minus Fairport is shown in Figure 6-5. The rate of vertical movement as determined from the slope of the linear regression line is -21.81 cm/century, indicating that Fairport is falling relative to Buffalo. However, it has been brought to the attention of the Coordinating Committee that both the water level gauge and the benchmark at Fairport are subsiding. It has been suggested that this may be the result of salt mining in the area. Regardless of the underlying cause, the rate of movement at Fairport is far greater than what would be expected when compared to other Lake Erie water level gauges. For these reasons, it is suggested that Fairport be omitted from further analysis of glacial isostatic adjustment rates. 53 Buffalo - Fairport 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1970 1980 1990 2000 2010 Year Figure 6-5: Buffalo minus Fairport water level difference plot. 6.7. Cleveland, Ohio The available water level record begins in 1860 at Cleveland; however, data collected up until November 1903 is staff gauge data (Coordinating Committee, 1987). It is evident in the water level difference plot for Buffalo minus Cleveland that the variability in the staff gauge data collected prior to 1903 is greater than the variability in the data collected with a recording gauge after this year (Figure 6-6). The rate of vertical movement as determined from the slope of the linear regression line for the full period of record is -9.01 cm/century, indicating that Cleveland is falling relative to Buffalo. Alternatively, if the staff gauge data is excluded and a linear regression line is fit to the post-1903 data only, the rate of apparent vertical movement is only -7.09 cm/century. Moreover, removal of what appears to be an outlier in 1905 causes the rate to decline slightly more to -6.69 cm/century. Since the Buffalo gauge was moved in 1911 (Coordinating Committee, 1987), removal of at least this outlier seems reasonable. 54 Buffalo - Cleveland 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (Post-1903) Water Level Difference (m) 0.2 0.1 0 -0.1 1840 1860 1880 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-6: Buffalo minus Cleveland water level difference plot. In addition, note that the Cleveland gauge was moved several times during the past century of data collection. Though it can not be said with any certainty, this may be partly the cause of what appears to be a stepwise pattern to the water level difference plot as shown in Figure 6-7. There appear to be a number of sudden drops, interspersed with longer periods of relative stability between the two gauges. While this impression may simply be the result of variability in the plot, it also lends further evidence to the suggestion that, despite the assumptions of this and past studies, the apparent vertical movement between gauge pairs may not be linear. 55 Buffalo - Cleveland 0.3 June to Sep. Avg. 5-year Centred Running Avg. Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-7: Buffalo minus Cleveland water level difference plot with 5-year centred running average. 6.8. Marblehead, Ohio The available water level record begins in 1959 at Marblehead. The water level difference plot for Buffalo minus Marblehead is shown in Figure 6-8. The rate of vertical movement as determined from the slope of the linear regression line is -6.84 cm/century, indicating that Marblehead is falling relative to Buffalo. There do not appear to be any significant discontinuities in the water level difference plot. 56 Buffalo - Marblehead 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-8: Buffalo minus Marblehead water level difference plot. 6.9. Toledo, Ohio The water level record at Toledo begins in 1877, but up until 1911 only a handful of measurements are available at this location. Only after 1911 does the water level record at Toledo become more or less continuously available. In addition, recording gauges were installed at Toledo beginning in 1940, with staff gauges used prior to this time (Coordinating Committee, 1987). Furthermore, until 1939, the gauge was located on the Maumee River approximately 6 kilometres upstream of the current location, which is still on the river, but near the mouth and Maumee Bay. Shorter relocations have also taken place at both of these main gauge locations. The water level difference plot for Buffalo minus Toledo is shown in Figure 6-9. The rate of vertical movement as determined from the slope of the linear regression line is -5.93 cm/century, indicating that Toledo is falling relative to Buffalo. If the pre-1940 staff gauge data is omitted, which is also the data collected prior to moving the gauge closer to Maumee Bay, the calculated rate is -6.76 cm/century. 57 Buffalo - Toledo 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1964+) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-9: Buffalo minus Toledo water level difference plot. A significant amount of variability can be observed in the water level difference plot of Buffalo minus Toledo. In addition, although the recording gauge data collected from 1940 on is more complete than the staff gauge data collected prior to this date, unlike some other gauges noted previously (e.g. Milwaukee vs. Harbor Beach), it is difficult to tell whether there is significantly more noise in the staff gauge data than in the recording gauge data. This could be the result the gauge being located on the Maumee River prior to 1940. Furthermore, Lake Erie, due primarily to its geographic orientation, size and depth, is known to be the Great Lake whose recorded water levels are most affected by climate-related events. For example, it is not uncommon for sustained heavy winds to cause water levels measured at one side of Lake Erie to be different from water levels measured at the opposite side by a metre or more. Water levels at Toledo and Buffalo, which are located on opposite ends of the lake, would therefore be highly affected by such climatic events. An example is shown in Figure 6-10, which compares plots of water level differences between Toledo and Fermi Power Plant with those of Toledo and Erie, PA. As can be seen, because of their close proximity, the Toledo minus Fermi Power Plant plot of water level differences shows less variability and noise than the plot of Toledo minus Erie, PA, which are located further apart. Though the months of June through September were chosen to try and minimize these issues, they cannot be avoided entirely, and this would partly explain the higher degree of variability seen in these water level difference plots than in those of other gauge pairs. 58 Toledo - Fermi Power Plant vs. Toledo - Erie, PA June to September Average 0.3 Toledo - Fermi P.P. Toledo - Erie, PA Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-10: Comparison of Toledo minus Fermi Power Plant and Toledo minus Erie, PA water level difference plots. The high degree of variability in the plot makes it difficult to be confident in the estimated rate of apparent vertical movement. The variability may also be responsible for what appears to be a change in slope, which according to breakpoint analysis, occurred around 1977 (Figure 6-11). Although only recording gauge data is shown in Figure 6-11, the breakpoint was also estimated at 1977 when staff gauge data was included. However, the gauge was relocated in 1964, with the intake located downstream approximately 50 feet from the previous location (Coordinating Committee, 1987). This is one possible cause for the apparent change in slope. The rate of apparent vertical movement calculated with only post-1964 data is -12.64 cm/century. In any case, the rate of apparent vertical movement between Buffalo and Toledo remains difficult to estimate due to noise in the data, coupled with a number of gauge relocations, and the change of data collection method from staff gauge to recording gauge readings. 59 Buffalo - Toledo 0.3 June to Sep. Avg. Piecewise Regression Estimated Breakpoint Water Level Difference (m) 0.2 0.1 1977 0 -0.1 1930 1940 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-11: Buffalo minus Toledo breakpoint estimation for recording gauge period. 6.10. Monroe, Michigan According to the Coordinating Committee (1987), water levels started being collected in 1859 at Monroe; however, water levels were only collected in fragmented intervals, with large gaps in between. More importantly, published water levels are now only available from September 1975 until November 1988. With only 13 years of complete data available, this is a very short period of record and likely far too short for estimating rates of apparent vertical movement at this location. Nevertheless, the water level difference plot for Buffalo minus Monroe is shown in Figure 6-12. The rate of vertical movement as determined from the slope of the linear regression line is -25.11 cm/century, indicating that Monroe is falling relative to Buffalo. However, rates calculated for other combinations of months were significantly different (for example, the rate calculated taking the annual average of all monthly data was -75.33 cm/century). Furthermore, both of these calculated rates far exceed those calculated for other gauges near Monroe, or anywhere else on Lake Erie for that matter, and these rates are well beyond those which can reasonably be expected due to glacial isostatic adjustment. The short period of record certainly affects the accuracy of the estimate, and unfortunately, the Monroe gauge has been discontinued, so it is unlikely that additional water level data will be added to the period of record. Since the calculated rate of apparent vertical movement appears questionable for these reasons, Monroe should be omitted from further analysis. 60 Buffalo - Monroe 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1970 1975 1980 1985 1990 Year Figure 6-12: Buffalo minus Monroe water level difference plot. 6.11. Fermi Power Plant, Michigan Water level data has been collected since 1963 at Fermi Power Plant. The water level difference plot for Buffalo minus Fermi Power Plant is shown in Figure 6-13. The rate of vertical movement as determined from the slope of the linear regression line is -10.10 cm/century, indicating that Fermi Power Plant is falling relative to Buffalo. This rate seems reasonable based on the location of this gauge, the calculated rates at other gauges on Lake Erie, and the adequate length of the period of record. Visual inspection of the plot does not reveal any significant discontinuities. 61 Buffalo - Fermi Power Plant 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 6-13: Buffalo minus Fermi Power Plant water level difference plot. 6.12. Bar Point, Ontario The water level record begins in 1966 at Bar Point. The water level difference plot for Buffalo minus Bar Points is shown in Figure 6-14. The rate of vertical movement as determined from the slope of the linear regression line is -18.64 cm/century, indicating that Bar Point is falling relative to Buffalo. However, this rate was greater than the rates calculated using other combinations of months, including that calculated taking the annual average of all monthly data, which had the slowest calculated rate of -15.01 cm/century. Regardless, these rates are all greater than that which would reasonably be expected due to glacial isostatic adjustment alone. These disparities are in part due to the differences in available data in any given year. Specifically, if data for a certain month is missing in a given year, then an average including that missing month cannot be calculated for that year, while an average that does not include the missing month can be calculated and is used in the regression analysis. For example, if for some reason only the month of June is missing in a certain year, then the four-month May to August and June to September averages can not be calculated for this year. On the other hand, the July to October four-month average could be calculated. A water level difference value for this year would therefore be included for July to October but not for the other two four-month averages, and if for some reason this year were to have a significant influence on the slope of the regression line, it would 62 cause a significant discrepancy between the calculated rates of vertical movement. To account for this, only concurrent data (i.e. only years where all combinations of monthly averages can be calculated) can be used. When this was done for Bar Point, the calculated rates of apparent vertical movement showed somewhat better agreement. Differences in the calculated rates, however, do still remain, and these remaining disparities are a reflection of the short period of record and the high degree of variability in the water level differences between Bar Point and Buffalo, which result at least in part due to meteorological-related factors similar to those described for Toledo above. Buffalo - Bar Point 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 6-14: Buffalo minus Bar Point water level difference plot. 6.13. Kingsville, Ontario The water level record begins in 1962 at Kingsville. The water level difference plot for Buffalo minus Kingsville is shown in Figure 6-15. The rate of vertical movement as determined from the slope of the linear regression line is -9.80 cm/century, indicating that Kingsville is falling relative to Buffalo. Despite the variability in the plot, this rate is similar to rates calculated for other combinations of months. Furthermore, there do not appear to be any significant discontinuities in the water level difference plot. 63 Buffalo - Kingsville 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 6-15: Buffalo minus Kingsville water level difference plot. 6.14. Erieau, Ontario The water level record begins in 1957 at Erieau. The water level difference plot for Buffalo minus Erieau is shown in Figure 6-16. The rate of vertical movement as determined from the slope of the linear regression line is -6.31 cm/century, indicating that Erieau is falling relative to Buffalo. The rate calculated for other combinations of months varies somewhat (for example, the rate calculated for the May to August average is -7.18 cm/century), again the result of variability in the plot. Otherwise, there do not appear to be any significant discontinuities in the water level difference plot. 64 Buffalo - Erieau 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-16: Buffalo minus Erieau water level difference plot. 6.15. Port Stanley, Ontario The water level record begins in 1908 at Port Stanley; however, data was collected only in the summer months until 1911, and no data was collected from November 1911 until collection was resumed in June 1926. The gauge was also relocated during this period. The water level difference plot for Buffalo minus Port Stanley is shown in Figure 6-17. The pre-1926 period clearly does not fit well with the remaining data. The rate of vertical movement as determined from the slope of the linear regression line is -2.14 cm/century if pre-1926 data is included, and -3.47 cm/century if only data from 1926 onwards is used. The latter rate seems to be the more acceptable of the two, especially for the current period. Either way, these rates indicate that Port Stanley, similar to nearly all other stations on Lake Erie, is falling relative to Buffalo. Also of note is that the Port Stanley gauge is susceptible to silting issues, and is scheduled for relocation in the near future; however, these issues are mainly a concern with instantaneous readings (Carol Robinson, personal communication, 2008). Silting issues should not, therefore, have a significant impact on the monthly mean data used in this study. 65 Buffalo - Port Stanley 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1926+) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-17: Buffalo minus Port Stanley water level difference plot. 6.16. Port Dover, Ontario The water level record begins in 1958 at Port Dover. The water level difference plot for Buffalo minus Port Dover is shown in Figure 6-18. The rate of vertical movement as determined from the slope of the linear regression line is -1.82 cm/century, indicating that Port Dover is falling relative to Buffalo. Other than a few periods of missing data, there do not appear to be any significant discontinuities in the water level difference plot. 66 Buffalo - Port Dover 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 6-18: Buffalo minus Port Dover water level difference plot. 6.17. Port Colborne, Ontario The water level record begins in 1911 at Port Colborne. The water level difference plot for Buffalo minus Port Colborne is shown in Figure 6-19. The rate of vertical movement as determined from the slope of the linear regression line is -5.11 cm/century, indicating that Port Colborne is falling relative to Buffalo. The 5-year centred average plot indicates what appears to be a stepped pattern to the water level differences (Figure 6-20). In addition to the stepped pattern, it appears that any apparent vertical movement occurring in the past has been nearly stopped or suspended since approximately 1979 (the rate calculated from 1979 onwards is 0.34 cm/century). With Buffalo and Port Colborne being located only approximately 30 kilometres apart, one would expect such a slow rate of apparent vertical movement between these two locations. In addition, their close proximity would suggest there should be less variability in the plot. While this does appear to be true after approximately 1979, it does not appear to be evident prior to this time. The observed stepped pattern is similar to that seen at Cleveland, and less so to a pattern seen at Port Stanley (Figure 6-21). The only other long-term water level gauge on Lake Erie is Toledo, but this gauge does not provide much additional information or an accurate comparison due to the issues at Toledo already discussed in Section 6.9. 67 Buffalo - Port Colborne 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-19: Buffalo minus Port Colborne water level difference plot. Buffalo - Port Colborne 0.3 June to Sep. Avg. 5-year Running Avg. Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-20: Buffalo minus Port Colborne water level difference plot with 5-year centred running average. 68 Buffalo - Lake Erie Gauges 5-year Centred Running Average 0.3 Buf.-Cleveland Buf.-Pt. Colborne Buf.-Pt. Stanley Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-21: Buffalo minus various Lake Erie gauges 5-year centred running averages of water level difference plots. There is little evidence suggesting why such a pattern exists. The Buffalo gauge has been located at the same location since 1911, while the Port Colborne gauge has been moved twice (Coordinating Committee, 1987); the first move occurred in 1925, when the gauge was moved approximately one kilometre from within the Old Welland Canal to closer to the lake itself; the second relocation occurred in 1964, but was only a short distance from the previous location. While the first relocation potentially explains what appears to be a step occurring around the same time, neither relocation explains what appears to be a more significant step occurring sometime between 1940 and 1960, nor do they explain why Port Colborne has gone from appearing to be falling relative to Buffalo to appearing to be relatively stable. Lastly, Figure 6-22 shows a comparison of Buffalo minus Port Dover and Buffalo minus Port Colborne water level difference plots. Port Colborne has a longer period of record than Port Dover, and the rate of movement calculated for the entire period of record at Port Colborne is much different than the rate calculated for the entire period of record at Port Dover, which only begins in 1958. After 1958, however, during which data was collected at both locations, the water level differences appear to be quite similar. Due to the relatively close proximity of the two gauges, this is expected, and it again illustrates the importance of the period of record length on the calculated rates of apparent vertical movement between gauge pairs. 69 Buffalo - Port Dover vs. Buffalo - Port Colborne 0.3 Port Dover (1958+) Port Colborne (1958+) Port Colborne (p-o-r) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 6-22: Comparison of Buffalo minus Port Dover and Buffalo minus Port Colborne water level difference plots. 6.18. Summary Similar to Lakes Michigan-Huron, water level difference plots between gauge pairs on Lake Erie are often complicated. While water level differences at some gauge pairs follow a mostly linear relationship, others, such as Cleveland and Port Colborne, appear to follow a more stepwise pattern. In addition, the meteorological impacts on Lake Erie lead to significant variability and noise in the water level difference plots. This variability is mostly the result of meteorological factors, such as wind events, which cause significant short-term differences in water levels between different areas of the lake, in particular between the eastern and western ends. This variability increases the uncertainty of the slope estimates, and can also cause inconsistencies to be observed in the average water level difference plots. The variability can also potentially either hide or exaggerate inconsistencies caused by other possible factors, further complicating the analysis. A summary of the calculated rates of apparent vertical movement for Lake Erie are shown in Table 6-2. The uncertainty as given by the standard error of the slope value is also shown, as are the additional scenarios as described in previous sections. 70 Gauge (Period of Record) Rate Standard Error *Additional Scenarios (cm/century) (cm/century) Sturgeon Point (1969-2006) 1.14 0.79 Barcelona (1961-1987) -6.43 3.09 Erie, PA (1958-2006) -7.69 1.02 Fairport (1975-2006) -21.81 1.63 Cleveland (1860-2006) -9.01 0.36 -7.09 0.44 *Cleveland (1903-2006) -6.69 0.37 *Cleveland (1903-2006 & w/o 1905 outlier) Marblehead (1959-2006) -6.84 1.35 Toledo (1906-2006) -5.93 0.74 -6.76 1.19 *Toledo (1940-2006) -12.64 1.84 *Toledo (1964-2006) Monroe (1976-1988) -25.11 9.87 Fermi Power Plant (1964-2006) -10.10 1.72 Bar Point (1966-2006) -18.64 2.30 Kingsville (1962-2006) -9.80 1.74 Erieau (1958-2006) -6.31 1.34 Port Stanley (1909-2006) -2.14 0.65 -3.47 0.62 *Port Stanley (1926-2006) Port Dover (1958-2006) -1.82 0.91 Port Colborne (1912-2006) -5.11 0.38 0.34 1.18 *Port Colborne (1979-2006) Table 6-2: Summary of Lake Erie rates of apparent vertical movement and uncertainty relative to Buffalo. 7. Lake Ontario 7.1. Overview Water level data from a total of ten gauge stations were analyzed on Lake Ontario (Figure 7-1). Six of these gauge stations are located in Canada, while the remaining four gauge stations are located in the United States. Table 7-1 gives a list of the gauge stations used in this analysis, as well as a summary of the periods of record for each station. 71 Figure 7-1: Lake Ontario gauge locations. Station ID Location Start Year End Year Missing (Full) Years 1899-1913; 1915 -1908-34; 1936-52 -- Cape Vincent, NY, USA 1898 2006 Oswego, NY, USA 1860 2006 Rochester, NY, 9052058 1860 2006 USA 9052076 Olcott, NY, USA 1967 2006 Port Weller, ON, Canada 13030 1929 2006 1932-1955 Port Dalhousie, 13040 1910 1956 -ON, Canada Burlington, ON, Canada 13150 1970 2006 -Toronto, ON, 13320 1906 2006 -Canada Cobourg, ON, 13590 1956 2006 -Canada Kingston, ON, Canada 13988 1909 2006 -Table 7-1: Lake Ontario gauge stations and periods of record. 9052000 9052030 7.2. Number of Full Years Missing Total Years of Data 16 0 93 147 44 0 103 40 24 54 0 47 0 37 0 101 0 51 0 98 Cape Vincent, New York The Cape Vincent, NY, gauge was selected as the outlet of Lake Ontario. The Cape Vincent gauge is located closest to the outlet of Lake Ontario, it has traditionally been used as the outlet in past studies and the gauge has a relatively long period of record. Alternatively, the Kingston, ON, gauge also has a long period of record and is located close to both the outlet and the Cape Vincent gauge. The Kingston gauge was also investigated as a potential outlet in order to confirm or provide an alternative to results 72 obtained from Cape Vincent, although unfortunately there are a number of potential issues with the Kingston gauge that impact the results (Section 7.11). The Cape Vincent period of record begins in 1898; however, only four months of data are on record until 1916, after which point data is for the most part continuous. Also of note is that water level data collected between 1916 and 1935 is recording gauge data, data from 1936 until April 1955 is primarily staff gauge data (although May to November of 1954 were collected with a recording gauge), and following April 1955 all data is recording gauge data. Since 1916 the gauge has been relocated twice (1936 and 1939). 7.3. Oswego, New York The water level record begins in 1860 at Oswego. The water level difference plot for Cape Vincent minus Oswego is shown in Figure 7-2. The rate of vertical movement as determined from the slope of the linear regression line is -3.59 cm/century, indicating that Oswego is falling relative to Cape Vincent. There do not appear to be any significant discontinuities in the water level difference plot. Cape Vincent - Oswego 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 7-2: Cape Vincent minus Oswego water level difference plot. 73 7.4. Rochester, New York The water level record begins in 1860 at Rochester; however, from 1907 to 1953 no data is available other than five months worth in 1935. Furthermore, the data collected prior to 1907 is almost entirely staff gauge data, whereas the data collected after 1953 is from recording gauges (Coordinating Committee, 1987). The water level difference plot for Cape Vincent minus Rochester is shown in Figure 7-3. The pre-1953 data has been omitted from the analysis. The rate of vertical movement as determined from the slope of the linear regression line is -7.99 cm/century, indicating that Rochester is falling relative to Cape Vincent (the rate is -7.91 cm/century when the 1935 data is included). There do not appear to be any significant discontinuities in the water level difference plot. Cape Vincent - Rochester 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 7-3: Cape Vincent minus Rochester water level difference plot. 7.5. Olcott, New York The published water level record begins in 1967 at Olcott. The water level difference plot for Cape Vincent minus Olcott is shown in Figure 7-4. The rate of vertical movement as determined from the slope of the linear regression line is -12.52 cm/century, indicating that Oswego is falling relative to Cape Vincent. However, the rates calculated for other combinations of months differed fairly significantly. For example, the rates calculated for the annual, May to August and July to October averages were -8.80, -10.52 and -11.56 cm/century, respectively. This is at least partially the result of the relatively 74 short period of record at Olcott, as well as additional incomplete data within years. The annual average rate in particular was calculated with only 26 full years of data, whereas the June to September average was calculated with 37 values. The calculated rate of apparent vertical movement should become more certain in future years as more data is added to the period of record. Lastly, there do not appear to be any significant discontinuities in the water level difference plot. Cape Vincent - Olcott 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 7-4: Cape Vincent minus Olcott water level difference plot. 7.6. Port Weller, Ontario The water level record begins in 1929 at Port Weller; however, data is not available from November 1931 until January 1956, and the gauge was relocated during this time (Coordinating Committee, 1987). The water level difference plot for Cape Vincent minus Port Weller is shown in Figure 7-5. The rate of vertical movement as determined from the slope of the linear regression line is -14.32 cm/century if the pre-1956 data is included and -17.13 cm/century when only data from 1956 onwards is used. Either way, Port Weller appears to be falling relative to Cape Vincent. Since the gauge was relocated prior to 1956, and since the pre-1956 data significantly affects the calculated rate of apparent vertical movement, the rate calculated post-1956 is the more representative of the actual relationship between these two gauges. Lastly, there do not appear to be any significant discontinuities in the water level difference plot. 75 Cape Vincent - Port Weller 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Linear (1956+) Water Level Difference (m) 0.2 0.1 0 -0.1 1920 1940 1960 1980 2000 2020 Year Figure 7-5: Cape Vincent minus Port Weller water level difference plot. 7.7. Port Dalhousie, Ontario The water level record begins in 1910 but ends in 1956 at Port Dalhousie. The water level difference plot for Cape Vincent minus Port Dalhousie is shown in Figure 7-6. The rate of vertical movement as determined from the slope of the linear regression line is -15.06 cm/century, indicating that Port Dalhousie is falling relative to Cape Vincent. There appears to be a possible shift or step down in the plot around 1933, but since the period of record is not very long and ends in 1956, this was not investigated further. 76 Cape Vincent - Port Dalhousie 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1910 1920 1930 1940 1950 1960 Year Figure 7-6: Cape Vincent minus Port Dalhousie water level difference plot. 7.8. Burlington, Ontario The water level record begins in 1970 at Burlington. The water level difference plot for Cape Vincent minus Burlington is shown in Figure 7-7. The rate of vertical movement as determined from the slope of the linear regression line is -24.05 cm/century, indicating that Burlington is falling relative to Cape Vincent. Calculations of rates for other combinations of months yielded significantly different results, including a rate of -16.61 cm/century for the annual average (though only 19 years of full annual data are available) and -22.62 and -20.95 cm/century for the May to August and July to October averages, respectively. These discrepancies are a result of the short period of record at Burlington combined with missing data (a maximum of 33 years of data were available for the calculations for any monthly combination). In addition, the Burlington gauge was moved in 1981, and is now located on a concrete pier close to the sand bar and filled land that divides Lake Ontario and Hamilton Harbour. Tushingham (1992) noted that this gauge is potentially unstable for these reasons. There do not appear to be any significant discontinuities in the water level difference plot, but due to the reasons outlined the rate of apparent vertical movement calculated at Burlington should be treated with caution. 77 Cape Vincent - Burlington 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1960 1970 1980 1990 2000 2010 Year Figure 7-7: Cape Vincent minus Burlington water level difference plot. 7.9. Toronto, Ontario The water level record begins in 1906 at Toronto. The water level difference plot for Cape Vincent minus Toronto is shown in Figure 7-8. The rate of vertical movement as determined from the slope of the linear regression line is -10.93 cm/century, indicating that Toronto is falling relative to Cape Vincent. The water level difference plot appears to show a somewhat stepwise pattern, which is particularly evident in the 5-year centred running average plot (Figure 7-9). This is similar to some of the patterns seen on other lakes, such as at Calumet Harbor and at Milwaukee on Lake Michigan, for example (Sections 5.5 and 5.6, respectively). On Lake Ontario, however, a similar pattern is not as obvious at other long-term gauges. 78 Cape Vincent - Toronto 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 7-8: Cape Vincent minus Toronto water level difference plot. Cape Vincent - Toronto 0.3 June-Sep. Avg. June-Sep. 5-year Centred A Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 7-9: Cape Vincent minus Toronto water level difference plot with 5-year centred running average. 79 7.10. Cobourg, Ontario The water level record begins in 1956 at Cobourg. The water level difference plot for Cape Vincent minus Cobourg is shown in Figure 7-10. The rate of vertical movement as determined from the slope of the linear regression line is -8.51 cm/century, indicating that Cobourg is falling relative to Cape Vincent. There appears to be a downward shift or jump in the water level difference plot occurring in 1977, which is also concurrent with a gap in the data. However, while this apparent shift appears obvious in this plot, the same shift is not evident in the July to October plot, in which there is no data gap (Figure 7-11). The rate of vertical movement as calculated from the July to October average is -8.30 cm/century, which is close to that calculated from the June to September averages. Therefore, it seems likely that what appears to be a shift in the June to September plot is actually just an illusion caused by variability and the gap in the data in this case. This helps illustrate the difficulty in making assumptions of this sort based on a single plot of water level differences. Cape Vincent - Cobourg 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 7-10: Cape Vincent minus Cobourg water level difference plot (June to September average). 80 Cape Vincent - Cobourg 0.3 Jul-Oct Avg Linear (Jul-Oct Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1950 1960 1970 1980 1990 2000 2010 Year Figure 7-11: Cape Vincent minus Cobourg water level difference plot (July to October average). 7.11. Kingston, Ontario The water level record begins in 1909 at Kingston. The water level difference plot for Cape Vincent minus Kingston is shown in Figure 7-12. The rate of vertical movement as determined from the slope of the linear regression line is 3.03 cm/century, indicating that Kingston is rising relative to Cape Vincent. However, there appear to be a number of discontinuities in the water level difference plot. First, there appears to be a shift or plateau running from 1929 until 1941, at which point the water level differences shift down again. There is then a short rising period which continues until approximately 1954, after which the plot appears to level out or even descend slightly. Alternatively, the plot can instead be seen to rise until sometime in the late-1970s, after which it begins to descend slightly. The reasons for the odd shape of the plot are unknown. The apparent shift upwards occurring between 1929 and 1941 is also evident in a plot of Toronto versus Kingston, and somewhat less evident in a plot of Oswego versus Kingston, which would imply that Kingston data is the cause (Figure 7-13). Furthermore, though not major, a change in slope sometime in the 1970s or 1980s is also evident in these plots. A number of gauge relocations have taken place at Kingston, including in 1962, 1974 and 1977. The 1962 relocation in particular was a fairly significant one, as the gauge was moved more than 81 3 kilometres. The rate of apparent vertical movement calculated with data collected after 1962 is -2.85 cm/century. However, neither this date nor the other relocation dates appears to coincide well with any of the shifts or discontinuities occurring in the water level difference plot. Cape Vincent - Kingston 0.3 Jun-Sep Avg Linear (Jun-Sep Avg) Water Level Difference (m) 0.2 0.1 0 -0.1 1900 1920 1940 1960 1980 2000 2020 Year Figure 7-12: Cape Vincent minus Kingston water level difference plot. Since the Cape Vincent and Kingston gauges are only located approximately 16 kilometres apart, the expected rate of apparent vertical movement between the two locations should be small. The rate of 3.03 cm/century calculated seems high, especially when compared to other locations on Lake Ontario, and the differences between the rates calculated for them. For example, the difference in rate calculated for Oswego and Rochester is only 4.32 cm/century, despite being separated by approximately 90 kilometres, while the difference between Toronto and Olcott is only 1.59 cm/century, despite being separated by approximately 68 kilometres. 82 Kingston Water Level Difference Comparisons 0.1 Water Level Difference (m) 0 -0.1 Cape Vincent-Kingston -0.2 Toronto-Kingston Oswego-Kingston -0.3 1900 1920 1940 1960 1980 2000 2020 Year Figure 7-13: Comparison of Kingston water level difference plots. 7.12. Summary Water level difference plots on Lake Ontario are somewhat less complicated than those on Lakes Michigan-Huron and Lake Erie, but some issues are still evident, such as at Kingston, for example. Water level differences at most gauge pairs appear to follow a mostly linear relationship, though in addition to the complications at Kingston, Toronto appears to show a somewhat stepwise pattern similar to that seen on other lakes. A summary of the calculated rates of apparent vertical movement for Lake Ontario are shown in Table 7-2. The uncertainty (standard deviation of the slope value) is also shown, as are additional scenarios as described in previous sections. 83 Gauge (Period of Record) Rate Standard Error *Additional Scenarios (cm/century) (cm/century) Oswego (1916-2006) -3.59 0.28 Rochester (1935-2006) -7.91 0.58 -7.99 0.63 *Rochester (1953-2006) Olcott (1967-2006) -12.52 1.16 Port Weller (1930-2006) -14.32 1.05 -17.13 1.06 *Port Weller (1956-2006) Port Dalhousie (1916-1956) -15.06 1.33 Burlington (1970-2006) -24.05 1.70 Toronto (1916-2006) -10.93 0.42 Cobourg (1957-2006) -8.51 0.89 Kingston (1916-2006) 3.03 0.36 -2.85 0.74 *Kingston (1962-2006) Table 7-2: Summary of Lake Ontario rates of apparent vertical movement and uncertainty relative to Cape Vincent. 8. Conclusions This analysis has first and foremost provided an update to the rates of apparent vertical movement (believed to be caused primarily by glacial isostatic adjustment) as calculated from the slope of the linear regression line determined from water level differences between gauge pairs. The rates of movement and their associated uncertainty have been calculated between all gauges on a given lake and that lake’s outlet with all published data collected up to and including 2006. In addition to the revised rates, the key findings of this study are as follows: • The four-month June to September average water level differences should be used in this type of analysis. • The length of the period of record is important. Shorter periods of record (usually less than 40 years, approximately) are inadequate and should be omitted. • Glacial isostatic adjustment may not be linear in nature. Discontinuities, shifts, stepwise patterns, changes in rate and other interesting features can be seen in some water level difference plots. In some cases an explanation for these features can be suggested, whereas in other cases no explanation is immediately evident. Furthermore, in some instances a feature seen in one plot can be contradicted by its absence in another plot. • Water level difference plots can be helpful in identifying issues in the water level data itself. 84 • Meteorological effects can have an impact. • The rate of movement calculated for any given period of time may not be representative of the rate of movement over the entire period of record, or during any other time. • In future glacial isostatic adjustment studies, Lakeport, Michigan, should be considered for use as the outlet of Lakes Michigan-Huron as opposed to Harbor Beach, since Lakeport is located closer to the outlet, there appears to be movement between Lakeport and Harbor Beach, and since the period of record at Lakeport is now greater than 50 years. Water level difference plots can show significant variability. This has normally been attributed mostly to meteorological related factors. While this may be valid, it has also been shown that the data collection method (i.e. staff gauge versus recording gauge data) can influence the amount of observed variability. This variability increases the uncertainty with which the rates of vertical movement can be calculated. By using the four-month June to September average water level differences, the variability is reduced to some degree without substantially increasing the possibility of having missing data. In addition, the results are comparable to those of many past studies, which also used the June to September average. The length of the period of record used is important. Shorter periods of record tended to show rates of movement that were inconsistent with the rates calculated at other gauges on a given lake, and sometimes these rates were beyond the rates of movement that might be reasonably expected due to glacial isostatic adjustment. It has also been shown in a number of cases that sometimes the assumption of a linear relationship is invalid. Instead, water level difference plots can show a stepped pattern (e.g. Harbor Beach minus Milwaukee), a change in the rate of movement (e.g. Harbor Beach minus Goderich), or other inconsistencies that do not follow any obvious pattern (e.g. Cape Vincent minus Kingston). This is not the first study to report observed data issues. The Coordinating Committee (2001) report, for example, stated that systematic trends in outliers at a number of gauges should be investigated further. These trends could be the result of any number of issues outlined here. Determining the exact cause of these observed inconsistencies is beyond the scope of this document, and may actually be impossible to determine for certain in most cases. The observed non-linear patterns could be the result of any number of factors or a combination of them. Possible explanations include data collection method changes, meteorological impacts, gauge relocations or even actual geologic movements. For the most part the assumption of a linear relationship may be valid, but in reality glacial isostatic adjustment may not always be linear in nature. Perhaps in some areas the land remains relatively stable for periods of time, followed by short periods of faster movement, which would help explain the stepped patterns observed between some gauge pairs. On the other hand, the stepped pattern could instead simply be the result of a 85 cyclical meteorological impact. For example, perhaps the average wind conditions on a given lake vary in a multi-year, cyclical pattern (i.e. periods of stronger winds on average lasting for several years followed by periods of weaker winds). Since high winds can cause wind setup such that water levels at one end of the lake differ significantly from water levels at the other end, a cyclical difference in wind patterns would not have to be significant to cause a noticeable stepped pattern in the average water level difference plots, especially when one considers that the average water level differences themselves are measured in centimetres, whereas wind setup can cause short-term water level differences of a metre or more. Furthermore, perhaps the lakes themselves differ in how they are recovering from the weight of the glaciers. For the most part, water level differences on Lake Superior appear to follow a linear relationship, whereas discrepancies were observed more often on the other lakes. This could be the result of much of Lake Superior being located within the Canadian Shield, whereas the other lakes are not, or perhaps similar observations are only hidden on Lake Superior because it is generally recovering at a faster rate than the other lakes. Regardless, there is obviously still a lot we do not understand about glacial isostatic adjustment in the Great Lakes region. 9. Recommendations It should be noted that the importance of these issues is likely to vary depending on the problem being investigated. These issues may be less important in a study requiring analysis of long-term impacts of apparent vertical movement than in a study requiring analysis of short-term impacts. For example, to explain the land-to-water relationship around a lake and how the observed water levels at each location have changed over the past century, the average rate of movement calculated for the gauges using linear regression is likely adequate, regardless of whether the relationship is truly linear or not. Similarly, the volumes of each lake are believed to be changing with time, such that each lake is either storing or decanting water due to glacial isostatic adjustment. The magnitude of these changes, however, likely requires that long-term average rates of change be used in most cases. However, these issues can have more serious consequences if shorter periods are being investigated. For example, water level transfers are sometimes used to replace missing data at a certain gauge station. The measured water levels from one gauge are transferred to the other gauge to replace the missing data by using an estimated rate of vertical movement between the two gauges. In such cases, a long-term linear relationship between these gauges may not accurately describe the real relationship occurring during the period of missing data. For example, if the relationship is actually stepwise, as opposed to linear, the rate of vertical movement during the period in question may be significantly different than the long-term linearly calculated rate. The transferred data might be sufficient for some if not most purposes, but may be entirely inadequate for 86 other uses (such as determining the rate of apparent vertical movement between gauge pairs, for example). As another example, rates of apparent vertical movement are also being investigated using Global Positioning Systems (GPS). In these studies, GPS are used to measure rates of apparent vertical movement, and this is often done over far shorter periods of time than those studies that use water level differences. Again, the rate of apparent vertical movement measured over a shorter period may not be representative of the rate occurring over a longer period. On the other hand, GPS studies might be able to help support or refute some of the issues and discrepancies documented in this report. All things considered, it is advised that the rates of apparent vertical movement as calculated from linear regression of the water level difference relationships be used with caution depending on the analysis being undertaken. 87 10. References Coordinating Committee (1977). Apparent vertical movement over the Great Lakes. Report by the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data. July 1977. Coordinating Committee (1978). History of water level gauges: Upper Great Lakes and the St. Clair – Detroit Rivers. Report by the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data. January 1978. Coordinating Committee (1987). History of water level gauges: Lower Great Lakes and the International Section of the St. Lawrence River. Report by the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data. March 1987. Coordinating Committee (2001). Apparent vertical movement over the Great Lakes Revisited. Report by the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data. November 2001. Coordinating Committee (2006). 81st meeting minutes of the Coordinating Committee on Great Lakes Basic Hydraulic and Hydrologic Data. November 14-15, 2006. Mainville, A. and Craymer, M. R. (2005). Present-day tilting of the Great Lakes region based on water level gauges. Geologic Society of America (GSA) Bulletin. 117(7/8), July/August 2005, p. 1070-1080. Quinn, F. H. and Southam, C. (2008). Lake Huron water level gage analysis. Report for the International Upper Great Lakes Study. Tushingham, A.M. (1992). Postglacial uplift predictions and historical water levels of the Great Lakes. Journal of Great Lakes Research. 18(3), p. 440-455. 88
© Copyright 2026 Paperzz