INTRODUCTION

Since 1679, when the first commercial vessel was launched in the Great Lakes and subsequently disappeared, shipwrecks have been a common event. However, only recently has the public's interest in shipwrecks spread beyond a few amateur historians and divers. This increase in public awareness has created a large volume of new publications. Many offer information such as type of ships, dates, cargoes and wreck locations. Divers have also noted the importance of location as a primary concern. Robert Marx, an expert on Spanish shipwrecks in the Caribbean, explains:

There are people who erroneously believe that only a limited amount of research is necessary to locate an old shipwreck, because with today's fantastic underwater electronic equipment, any wreck can be located if only the general location of a site is known. A well known mineral company decided to enter the treasure hunting field. Using a deep submersible vehicle and the latest in electronic gear, it announced that it would make the biggest treasure recovery in history. Three months later, with over a quarter of a million dollars wasted, they were begging for my research (Marx, 15).

DATA SOURCES

Data was collected and compiled from a variety of sources; Shipwrecks of the Great Lakes by Dana Thomas Bowen; Shipwrecks of Lake Huron by Jack Parker; Lake Superior's Shipwreck Coast by Frederick Stonehouse; Sail and Rail by Lawrence Wakefield; Shipwrecks of the Sleeping Bear by Steve Harold; Directory of Shipwrecks of the Great Lakes by Karl Heden; Shipping Casualties Resulting in Total Loss on the Great Lakes From 1870 to 1970 by A. Winkelmann; Marine Disasters on the Western Lakes, during the Navigation of 1869 by Captain J.W. Hall; United States Department of Agriculture Report of Wrecks which occurred on the Great Lakes from December 17, 1885 to November 15, 1893; United States Coast Guard Report of Life Saving Service Reports of Assistance Rendered. These sources were chosen due to the significant amount of information listed on each shipwreck and the number of shipwrecks listed.

METHODOLOGY

The first step was to compile the information from the various data sources. A database software was written to compile each shipwreck entry into seventeen distinct variables such as date, cargo and lives lost. A random sample of twenty-four hundred shipwrecks was taken from the selected sources. Since the sampling method used replacement, the software then removed any duplicate entries. A duplicate entry was considered to be any shipwreck entry with matching names, type, year, and lake variables. This criteria removed over one hundred duplicate entries, but also removed any ships that happened to wreck twice in the same lake in the same year. After duplicate entries were removed, the sample data was processed into several data files. Each file produced a table in which two variables could be tested in a chi-square test. Other variables that could not be tested in a chi-square test were instead tabulated and mapped. Since any of the factors might have a relationship to location, they must be individually discussed.

Of these factors, the first to be considered is the physical landscape of the Great Lakes. Although the Great Lakes comprise over ninety-five thousand square miles of water, they offer little maneuvering room for the ships that sail them. Much of the Great Lakes contain dangerous shoals and reefs. Many of these were quickly mapped on navigation charts, but not all. For example, Superior Shoal, a shoal in Lake Superior that lies about 18 miles off the main route from the Soo to Thunder Bay, was not discovered until 1930. The areas containing large shoals or reefs is presented in figure 1.1

The Great Lakes also consist of minor waterways such as the Detroit waterway including Saint Clair River, Lake Saint Clair, and the Detroit River. Other narrow passages such as Saint Mary's River and the Niagara River connect the Great Lakes. These minor waterways are major congestion points of traffic. John Halsey, the State Archeologist for the State of Michigan, offers reasons why these waterways are considered dangerous.

Vessels often collided in these channels when navigation signals were misunderstood or improperly given, when the weather conditions were poor, or when one or more vessels traveled at an excessive speed...(Halsey, 22).

Since these hazards could have a direct influence on where shipwrecks might occur, then the number of shipwrecks located in these hazardous areas should be higher than other locations which are less dangerous. Figure 1.2 shows all locations which have over thirty shipwrecks in the general location.

Using both maps for comparison, there is a strong correlation between hazardous locations and the number of shipwrecks that occur. Areas such as Whitefish Point, Oswego, and Chicago had large numbers of shipwrecks due to the large amounts of traffic. Also, locations around the Manitou Islands, Presque Isle, Pointe Aux Barques had large number of shipwrecks due to the shoals and reefs nearby. However, not all areas that are congestion points or have shoals have large amounts of shipwrecks. The Keeweenaw Peninsula in Lake Superior is surrounded by the sawtooth reef, but has few shipwrecks. Cleveland is a major harbor for Great Lake shipping, but also has few shipwrecks. Therefore, other factors must have a greater influence on shipwreck location than just the physical landscape alone.

The Great Lakes have had over thirty different types of ships sailing across their waters. Some types such as barkentines and brigantines were introduced in the early 1800s, but were replaced by schooners, which could sail across the wind. Schooners were promptly replaced as soon as the steam engine was introduced. Steam engines powered the new vessels which appeared such as side-wheelers, propellers, tugs, lighters, dredges, car ferries, and whalebacks. All of these ships generally served one or two purposes such as hauling passengers, iron ore, railcars and other commodities across the lakes. To simplify the task of finding if there is a relationship between the type of ship and its location, the types of ships were broken down into 7 categories.

These categories are Tugs; Barges, including schooner-barges and steam barges; Steamers, which include propellers, side-wheelers, whalebacks, and all vessels that were driven by steam powered engines; Schooners, all ships that were rigged with a gaffed topsail and one or more headsails; Other Sailing Ships, which included vessels with squared rigged sails, barkentines, brigs, sloops, and sail yachts of substantial size; Vessels which included modern ships such as freighters, tankers, dredges, lighters, and any ship powered by oil or diesel powered engines; Not Available, since not all reference materials list the type of ship that wrecked.

These types are then compared to the Great Lake in which the wreck occurred. A chi-square test was used to determine if the type of ship is statistically independent of the lake in which the wreck occurred. The following three assumptions were made: that the observed and expected frequencies are distributed as a chi-square variable, an independent sample of size n=2240, and a significance level of .95. The rejection region was calculated using a confidence interval of .95 and degrees of freedom of 42. The resulting value of the region was 28.14 and the value of the chi-square was -195. Therefore, the conclusion can be made that there is no statistical relationship between type of ship and the lake in which the wreck occurred (See Appendix A for Calculations).

After the first shipwreck in 1679 of LaSalle's Griffon to the latest shipwreck of the Coast Guard Cutter Mesquite in 1992, the number of wrecks have been recorded by the year in which it wrecked. Since production centers, trade routes and type of ships change over time, so it is possible to hypothesize that the shipwreck locations change from one area to another over time also.

To compare time versus shipwreck location, it would be difficult to notice any trends by comparing every location to every year in which shipwrecks occured. However, by looking at decades and the lake in which the wreck occurred. it is possible to find a relationship between time and location. Since there are few shipwrecks before 1830 and after 1960, the category 1830 includes all wrecks before 1830; likewise, the 1960 category includes all wrecks from 1960 to present.

A chi-square test was used to determine if the lake where a wreck occurs is statistically independent or dependant of the decade in which the wreck occurred. The following three assumptions were made: that the observed and expected frequencies are distributed as a chi-square variable, an independent sample of size n=2420, and a significance level of .95. The rejection region was calculated using a confidence interval of .95 and degrees of freedom of 91. The resulting value of the region was 51.74 and the value of the chi-square was 545.82. Therefore, the conclusion can be made that there is a statistical dependency between the lake which a shipwreck occured and the decaded it occurred (See Appendix B for Calculations).

The type of accident refers to the type of shipwreck that occurs. Since not all shipwrecks result in a sunken ship, it is necessary to examine various categories of a wreck. These categories include founderings, strandings, capsizings, burnings, collisions and explosions. The significance of looking at different type of accidents lies with locating a shipwreck. According to Robert Marx, over 98 percent of shipwreck occuring in the Western Hemisphere occur in shallow waters. He explains that other 2 percent lost were due to ships capsizing in storms, fires at sea, ships sunk in battles (Marx 46). By looking at the type of accident, its possible to establish the general location of a shipwreck.

A chi-square test was used to determine if the lake where a wreck occurs is statistically independent or dependent of the type of accident. The following three assumptions were made; that the observed and expected frequencies are distributed as a chi-square variable, an independent sample of size n=2176, and a significance level of .05. The rejection region was calculated using a confidence interval of .95 and degrees of freedom of 56. The resulting value of the region was 39.8 and the value of the chi-square was 499.82. Therefore, the conclusion can be made that there is a statistical dependency between the lake which a shipwreck occured and the type of accident (See Appendix C for Calculations).

The total population of shipwrecks that have occured in the Great Lakes has been estimated from six thousand to over fifteen thousand. Although many historians and maritime researchers have documented many wrecks, only three to four thousand shipwrecks have been documented with its location. The remainder of shipwrecks have no documented source of its location.

This paper examined other documented factors such as the physical landscape, the type of ship, time, and the type of accident to discover if any relationships exist between these factors to a shipwreck's location. By using a database software, a sample of shipwrecks was used to examine each factor. Using maps and statistics, the following conclusions can be made. The general landscape indicates many hazardous areas, but these areas do not always contain large quanities of shipwrecks. The type of ship has no statistical relationship to a shipwreck's location. Therefore, a schooner or steamer have an equall chance of wrecking in any given location on the Great Lakes. However, time and type of accident does have a statistical relationship with shipwreck location. For example, it might be possible to discover a shipwreck site if the date and type of accident is known.