Cloquet and Moose Lake Fires 1918: Wind, Evacuation and Conflicting Totals

Historical wildfire analysis · Minnesota, United States · October 1918
Sources reviewed: 25 October 2026.

The Cloquet, Duluth and Moose Lake fires were a regional outbreak, not one flame front travelling through every affected community. On 12 October 1918, dry conditions and strong winds accelerated fires that had started earlier. Understanding that geography helps explain why a timeline for Cloquet cannot simply be applied to Moose Lake, and why regional loss figures require careful attribution.

Several origins, one destructive day

Minnesota Historical Society’s MNopedia places the origin of the Moose Lake fire near Tamarack on 4 October, and the main Cloquet–Duluth ignition near Brookston on 10 October. Its account describes rapid expansion on 12 October as a front brought stronger winds and falling humidity. The Cloquet–Duluth fire developed along separate eastward and southeastward paths. [1]

Historical stage	Event	Why it matters
4 October	Ignition associated with the later Moose Lake fire.	The destructive run was not the ignition date.
10 October	Ignition associated with the later Cloquet–Duluth fire.	A separate origin and development history.
12 October	Major expansion across several communities.	Multiple fronts affected different routes and destinations.
13 October	Statewide relief efforts began.	Emergency response continued after the main runs.

Chronology follows MNopedia. [1]

Wind observations need their averaging period

The National Weather Service retrospective reports a Weather Bureau peak wind of 76 mph at 17:52 and a five-minute wind speed of 65 mph. It also describes multiple fires across northeastern Minnesota. [2]

Those two wind values are not competing estimates of the same statistic. A short-lived peak and a five-minute value describe different temporal scales. Neither should be assigned to every location along the fire perimeter. A reconstruction needs the observation site and instrument context before transferring the measurements into a model.

Likewise, wind speed is not fire-spread speed. Dividing the distance between towns by 76 mph would not produce a defensible fire-arrival time. Fire movement depends on fuel, terrain, ignitions ahead of the front and changing local conditions, none of which is represented by that arithmetic alone.

Evacuation outcomes varied between communities

MNopedia describes organized warning and four evacuation trains in Cloquet, with more than seven thousand people leaving by rail. Its Moose Lake account includes rescue trains and refuge in Moose Head Lake. [1] These histories show different combinations of warning, transport and exposure.

They do not establish a controlled comparison of evacuation methods. A lower death toll in one town cannot be attributed to rail transport alone without accounting for warning time, the population present, available routes and the fire’s arrival. Survivor descriptions of refuge also remain descriptions of that event, not a general instruction for current emergencies.

The death-toll discrepancy should remain visible

Institutional source	Published fatality total
Minnesota DNR forestry timeline	453
Minnesota Digital Library exhibit	453
NWS anniversary retrospective	1,000

The figures above appear in the cited sources as reviewed. [2] [3] [4] This article uses 453 as the attributed Minnesota DNR and Minnesota Digital Library total, while recording the conflicting NWS figure. It does not present either as a newly verified casualty census.

Averaging these numbers would be inappropriate: publication by separate institutions does not make them independent statistical measurements. Reconciliation requires the underlying lists, geographic coverage and counting conventions. The summaries alone do not establish why the NWS anniversary figure differs.

Burned area also depends on the event boundary

The Minnesota Digital Library exhibit describes more than fifty fires combining to burn approximately 1,500 square miles. The NWS anniversary page gives 250,000 acres. [4] [2] Converting 1,500 square miles at 640 acres per square mile gives 960,000 acres, which is substantially different.

This calculation flags a discrepancy rather than resolving it. The next research step would compare mapped boundaries and determine whether the figures refer to all fires, selected incidents or particular land categories. This article does not assume that changing units explains the difference.

How to build a useful reconstruction

Start with separate records for each ignition and progression path. Attach each observation to a place, time and source. Keep weather measurements distinct from reported fire behavior, and retain the geographic scope of losses. Only then combine records into a regional account.

The result should explain where the evidence agrees and where it remains unresolved. This retrospective reviews institutional summaries, including indexed exhibit text; it does not claim a complete reading of the 1921 relief commission report, a new perimeter reconstruction or an independent casualty audit.