NATURAL FIRE REGIMES and PRE-EUROPEAN SETTLEMENT VEGETATION of the CHARLES M. RUSSELL NATIONAL WILDLIFE REFUGE Box 110 333 Airpo

Home , James Willard Schultz, Sacagawea River

NATURAL FIRE REGIMES AND PRE-EUROPEAN SETTLEMENT VEGETATION OF THE CHARLES M. RUSSELL NATIONAL WILDLIFE REFUGE

Report Prepared for

The Charles M. Russell NWR U.S. Fish and Wildlife Service Box 110 333 Airport Road Lewistown, MT 59457

Cecil C. Frost Landscape Fire Ecologist 119 Potluck Farm Road Rougemont, NC 27572 336-364-1924 (office) 919-906-1915 (cell)

August 22, 2008

ACKNOWLEDGEMENTS

Thanks are due to the staff of the refuge at Lewistown, and managers of Sand Creek, Jordan and Fort Peck. Thanks especially to Bob Skinner as my chief contact for questions, for discussions and for copies of refuge documents and published materials. Thanks go to Joann Dullum for producing the vegetation map and a GIS map of original habitats of prairie dogs in the CMR landscape, to Susan Langley for production of the fire frequency map on GIS, to Paul Pallas for help with assembling crews with equipment for collecting tree sections and for supplying information on fire records at CMR, and to Dan Harrell for field assistance and information on historical numbers of livestock in Montana. The Missouri Tree Ring Laboratory analyzed the historical fire scars in the first sections I collected form fire-killed trees and Michael Stambaugh and his father Phil Stambaugh came from the laboratory to CMR in 2007 to collect additional specimens to complete the fire scar chronologies.

DISCLAIMER

The findings and opinions expressed herein represent the interpretations and professional judgments of the author. These are not necessarily representative of the policies or opinions of The U.S. Fish and Wildlife Service.

2 TABLE OF CONTENTS Page Executive summary ...... 7 1) INTRODUCTION ...... 8 Reading the fire landscape...... 8 Field data ...... 9 Historical vegetation ...... 7 Changes in vegetation since European settlement ...... 10 Pre-European fire regimes ...... 10 Effects of the 2006 wildfires on old-growth Douglas fir and ponderosa pine ...... 10 2) FIRE ECOLOGY OF THE PHYSICAL LANDSCAPE ...... 13 Firebreaks and fire filters...... 14 Geologic features related to firebreaks and fire filters ...... 14 Firebreaks ...... 16 Fire filters ...... 16 Fire frequency versus fire severity ...... 17 Native Americans as an ignition source ...... 18 Effects of topography on fire frequency ...... 20 Lightning as an ignition source ...... 23 Ignition sources and fire movement ...... 23 Lightning ignition fans ...... 24 27 year fire history – the actual distribution of ignitions ...... 29 Implications for fire elsewhere in shortgrass prairie ...... 32 3) FIRE RELATIONS OF NATIVE ANIMAL SPECIES ...... 33 Historical effects of beaver on fire frequency, species diversity and possible habitat for moose at CMR ...... 33 Possible relationship between extirpation of beaver and decline of habitat for sharptail grouse ...... 34 History of beaver trapping in the CMR area ...... 35 Diversion of small stream water supplies for irrigation ...... 39 Possible moose habitat maintained by beaver ...... 39 Effects of prairie dogs on fire frequency ...... 40 Mapping historical extent of prairie dogs towns as step in fire frequency mapping ...... 42 The reported historical prairie dog increase ...... 44 Effects of bison on fire frequency ...... 45

3 Bison numbers ...... 45 History of cattle and sheep grazing in the CMR area ...... 46 The era of the big ranches 1880-1910 ...... 47 Effects of cattle and sheep grazing on fire frequency ...... 51 Fire spread in grazed and overgrazed prairie ...... 51 Effects of cattle ...... 52 4) TIMBER USE AND FIRE ...... 56 5) FIRE SCAR CHRONOLOGIES ...... 57 Considerations in using fire scar chronologies for determining historical fire frequency ...... 57 Suggestion of a higher fire frequency in the pre-horse era ...... 59 Buffalo grazing and fire – evidence from the fire scar record ...... 60 6) EVIDENCE FROM TREE DEMOGRAPHY PLOTS ...... 62 Lost Creek ...... 62 C.K. Creek ...... 64 7) EVIDENCE FROM HISTORICAL PHOTOS AND OTHER FIRE HISTORY ...... 65 8) MAPPING PRE-EUROPEAN FIRE FREQUENCY IN SHORTGRASS PRAIRIE ...... 70 Fire exposed versus fire sheltered sites ...... 72 Some steps in mapping both pre-European settlement fire frequency and vegetation ...... 72 Making a relative fire frequency map ...... 72 Drawing fire compartment boundaries ...... 73 Assigning fire frequency classes to fire compartments ...... 74 Assigning fire frequency numbers to each fire frequency class ...... 76 Fire frequency classes – variation between original fire frequency before and after introduction of the horse to North America by the Spanish ...... 77 Descriptions of the nine fire frequency classes ...... 78 Skewness in fire frequency distributions ...... 80 Fire frequency in the pre-horse era ...... 80 Considerations for using these tables for prescribed fire ...... 81 9) CONCLUSIONS: results products the Answer to the Question “Is this natural?” ...... 82 Speculative conclusions deserving further study...... 83 Arguments for natural versus anthropogenic causes of the severe fires of July 2006 ...... 83 LITERATURE CITED ...... 84 APPENDICES APPENDIX 1. HISTORICAL VEGETATION AND FUEL TYPES OF THE CHARLES M RUSSELL NATIONAL WILDLIFE REFUGE

4 APPENDIX 2. METHODS APPENDIX 3. CHRONOLOGY OF SELECTED HISTORICAL EVENTS AT CMR APPENDIX 4. POSSIBLE FUTURE WORK AT CMR

LIST OF TABLES Table Page 1. Historical vegetation and fuel types of the CMR ...... 9 2. Nine kinds of fire history evidence available at CMR ...... 10 3. Amount of local relief at various points around CMR ...... 20 4. Some animal-fuel-fire cascades ...... 33 5. Preferred soils, used more than once by prairie dogs ...... 43 6. Fire-return intervals, averaged between the three fire scar chronology sites ...... 61 7. Some kinds of evidence used in mapping original fire regimes ...... 71 8. Nine kinds of fire history evidence available at CMR ...... 71 9. Landscape factors influencing fire frequency ...... 75 10. Fire frequency classes at CMR 1730-1882 ...... 77 11. Fire frequency classes at CMR before introduction of the horse around 1730 ...... 81 12. Summary of results ...... 82

LIST OF FIGURES Figure Page 1. Western end of study plot CMR04 after fire ...... 11 2. Figure 1 before fire ...... 11 3. Eastern end of study plot after fire ...... 12 4. Figure 3 before fire ...... 12 5. Location of study plot CMR04...... 13 6 Unvegetated Bearpaw Shale ...... 15 7. A landscape-scale firebreak ...... 16 8. Fire filter soils near Hell Creek ...... 16 9. Badlands on the north side of Haxby Road ...... 17 10 Topographic roughness model for CMR ...... 21 11 Study plot on a clay soil in valley County dominated by Atriplex dioica ...... 22 12 Lightning fire at Middle Eighth Ridge ...... 23 13 The Little Rocky Mountains generating lightning ...... 25 14 The Little Rockies ignition fan ...... 26 15 The resorption phase of am orographic storm cloud ...... 26

5 16 The Judith Mountains ignition fan ...... 27 17 The Snowy Mountains ignition fan ...... 28 18 All four ignition fans superimposed...... 29 19 Lightning strike density of the U.S...... 29 20 Fires in the eastern third of CMR 1980-2007 ...... 30 21 Fires in the western third of CMR 1980-2007 ...... 31 22 Fires in the middle third of CMR 1980-2007 ...... 32 23 A natural beaver ecosystem retaining water in summer on upper Armells Creek ...... 35 24 An altered beaver system with a small beaver impoundment ...... 39 25 Prairie dog town on southernmost upland flat of U.L. Bend ...... 41 26 Bison on the Theodore National Wildlife Refuge ...... 45 27 Loading wool for market from James Fergus’ ranch at Armells in the 1890s ...... 48 28 Sheep on overgrazed winter range in the Yellowstone breaks, 1907 ...... 49 29 August 2007 fire in overgrazed shortgrass prairie ...... 52 30 30a through f: Graph of livestock numbers in Montana 1887-2006 ...... 53 31 Fire scar chronology for Sand Creek ...... 57 32 Fire scar chronology for Lost Creek ...... 59 33 Fire scar chronology for Soda Creek ...... 60 34 Cattle, sheep and horses on open range in Montana 1867-1886 ...... 61 35 Livestock numbers in Montana from 1887-1910 ...... 62 36 Lost Creek demography plot ...... 63 37 100% mortality in a ponderosa pine stand at Lost Creek killed by fire ...... 63 38 C.K. Creek demography plot ...... 64 39 C.K. Creek site ...... 64 40 Grassland on way down to Hell Creek, 1902 ...... 65 41 Repeat photos by Shantz, Phllips and Kay with vegetation change in the Judith Mountains 66 42 August 2007 fire along Musselshell Trail ...... 67 43 Photo taken by F.J. Haynes, 1880 at the mouth of Cow Creek ...... 68 44 Matching photo at Cow creek taken August 2007 ...... 69 45 Close-up of the portion of the ridge seen on the right side in the 1880 photo ...... 69 46 Another photo taken at Cow Creek 1880, looking north ...... 70 47 2007 photo to match Figure 46 ...... 70 48 Sample portion of the completed fire frequency map ...... 74 49 Juniperus horizontalis-J. communis-J. scopulorum, fire frequency indicator community ... 77

EXECUTIVE SUMMARY

Project Title: PRESETTLEMENT VEGETATION AND NATURAL FIRE REGIMES OF CHARFLES M. RUSSELL NATIONAL WILDLIFE REFUGE

PRINCIPAL INVESTIGATOR: Cecil Frost

The goal of this project was to develop maps of the historical fire regimes and original vegetation of CMR to provide a background for decisions around restoration and land management with fire. The intent was to produce the best approximation of the natural fire regimes and vegetation that existed at time of the demise of the buffalo in 1882. This is the vegetation that dominated the landscape prior to that time for some 6,000 years. Understanding these is essential to restoring habitats and managing lands for the full range of animal and plant species that depend upon them for habitat. Nearly all the upland original vegetation of the CMR was in some way structured by fire. The fire pattern at CMR is complex: some areas were exposed to frequent fire while others were very fire sheltered. Since the site contains many species that need some degree of fire or shelter from fire, the GIS layers can serve as base maps for guidance in using fire to restore habitats. Potential also may exist for restoring some beaver wetlands and other natural vegetation communities and wildlife habitats originally present and now diminished or extirpated. In the interest of preventing loss of old-growth trees, such as happened in the July 2006 fires, it may be possible to manage fuel loads in the more fire-exposed parts of the refuge such as the Lost Creek area, and in prairie-woodland boundaries, such as those in the Sand Creek section, to prevent stand-destroying fires that would have not been typical of the natural fire intensity in those places. A new mapping method using landscape fire ecology was used to reconstruct presettlement fire frequency (Map 1) and presettlement vegetation (Map 2). This involved using nine lines of evidence to reconstruct original conditions. These included fire scar chronologies from three different places on the refuge; field sampling of the best remnant vegetation on many of the 282 soil mapping units of the six counties of CMR; compilation of historical photos and other historical information relating to vegetation and fire; estimating fire effects in each kind of vegetation on each soil series and each topographic situation; characterizing things that contribute to ignitions and promote the spread of fire versus those that obstruct fire in the landscape; mapping regional and local fire compartments; and identification of fire-frequency indicator species and fire-frequency indicator plant communities. Descriptions were prepared of the original vegetation and fuel types of the refuge for each soil series. The presettlement fire regime method used here is expected to have application throughout the West in landscapes where frequent fire was an important determinant of vegetation in presettlement times. Such maps can be used as guides for to meet NWR wildlife and recreation objectives while restoring natural fire regimes and maintaining examples of the full range of rich natural communities and species that the area first encompassed. Significant findings: o The pre-European fire regime at CMR was lightning driven, in contrast with some other parts of the country where Native Americans were the dominant force. This was true at least after the 1730-1750 acquisition of horses by the Blackfeet. Prior to 1730 the Blackfeet and their precursors along the upper Missouri River may have used fire in the landscape in ways similar to those of the Mandans in the early European contact period downstream. If true, that would indicate that the fire frequencies indicated on the GIS map are conservative. o Ignitions are more frequent at CMR than in many other parts of the shortgrass prairie because of four local orographic ignition sources: the Little Rocky Mountains, the Judith Mountains, the

7 Bearpaw Mountains and the Snowy Mountains. Ignitions from the four sources reach their maximum in the center of the refuge and this pattern has existed for thousands of years. Any burning by First Nations people would have been superimposed on this fixed lightning ignition pattern. o Fire frequency is lowest in the eastern third of CMR, where fire spread is limited by low ignition frequency, extensive fire filters and firebreaks created by poorly vegetated gumbo and badland soils. o Lightning ignition density overrides topography in control of fire frequency in the western and central portions of CMR. o Global warming is known to be having an effect on fire frequency (Westerling et al. 2006): Climatic warming increases the length of the fire season resulting in more fires per year. But the killing of old-growth trees by fires at CMR is the result of increased fire intensity. The fire scar chronologies and tree demography plots indicate that woody fuels increased greatly beginning with saturation of the landscape by sheep and cattle around 1900. The replacement of bison with cattle and sheep led to a doubling of the length of fire intervals in the first half of the 20th century, providing enough time for accumulation of lethal amounts of juniper, as well as saplings of pine and fir that would have been kept down under the natural fire frequency. o A startling and unexpected finding was the impact of bison grazing on presettlement fire frequency in shortgrass prairie revealed by the three fire scar chronologies. Fire frequency could have been four times higher without buffalo. o The fire scar data also documented that the replacement of bison with cattle and sheep doubled the of length of fire intervals in the 20th century (there is now half as much fire as there was in the time of the buffalo). This is almost certainly the key factor in the severity of recent fires.

INTRODUCTION

Shortgrass prairie and desert grasslands are probably the most complex ecosystems in North America in terms of natural fire frequency. Fuels are often discontinuous because of prairie dogs towns, rock or barren clay outcrops or soils too poorly productive of grasses to support fire. Gumbo soils derived from shale support almost no plant cover in places while in other places the only cover consists of spring annuals that are green until June, then dry out and are too short or sparse to serve as fuel.

Reconstructing fire frequency is easier in Great Plains tallgrass prairie, Missouri oak savannas and southeastern longleaf pine savannas because light, flashy grass fuels were once continuous over large areas. In the midgrass and tallgrass prairies the grass fuels were continuous across vast landscapes. Fire intervals as frequent as every one to three years are known from such regions. There are historical records of fires that started near the toe of the Colorado Rockies that burned all the away across two states to the Mississippi River before going out.

Reading the fire landscape. Understanding fire frequency requires, for any point on CMR, being able to interpret whether you are in a fire sheltered or fire exposed area; observing the sources of fire, which include both fires ignited nearby on the site or fires that had access to the site, perhaps from a source many miles distant, by way of fire paths composed of continuous grass fuels; the quantity and extent of fuels— primarily grasses or pine needles—and connectedness of fuel patches; the presence of any fire filters, the distance from firebreaks and a set of other factors which will be discussed below and summed up in section nine on mapping fire frequency. This constitutes a different way of looking at the landscape and one that is essential to understanding the distribution of plant and animal species.

All of these natural influences on fire frequency, behavior and intensity have implications for applying fire during prescribed burning.

8 Field data. A number of kinds of data were collected, including study plots using worksheets in the field. Several kinds of such data are not covered in this report but will be dealt with elsewhere. These include data from field worksheets used for defining prairie dog habitat in the CMR area; flora of prairie dog towns; data collected to estimate intensity of use by browsers and grazers in 1/10 hectare study plots; species lists and cover values for each species collected in the 1/10 hectare plots, and data collected on fire relations of big sagebrush. All of these need further work and the data collected was used as adjuncts to looking at the effects on natural fire flow in the landscape. Full analysis of these five kinds of data is beyond the scope of the funded fire frequency mapping project. The following report and appendices focus on pre-European fire regimes. the central issue for which the study was intended. Unless specified otherwise, everything that follows refers to conditions in the original, pre-European landscape.

HISTORICAL VEGETATION There were 21 vegetation types in seven major vegetation groups. (Table 1). Appendix 1 covers the 21 vegetation types on the GIS map of presettlement vegetation and fuel types of CMR. See the appendix for descriptions and photos of the vegetation types and a list of the soils on which each occurs.

TABLE 1. HISTORICAL VEGETATION AND FUEL TYPES OF THE CHARLES M RUSSELL NATIONAL WILDLIFE REFUGE GROUP 1. ROCK OUTCROPS, BADLANDS, SHALE AND GUMBO BARRENS 1.1 “Rock” Outcrops and Poorly Consolidated Geologic Sediments 1.2 Rock Outcrop-Vegetation Mosaic 1.3 Badlands 1.4 Dwarf Artemisia tridentata Flats-Dry Ridgetop Prairie Mosaic

GROUP2. PRAIRIES 2.1 Dry Needlegrass-Wheatgrass Prairie 2.2 Mesa-top Meadows 2.3 Dry Prairie-Juniper Shrubland Mosaic with Occasional Ponderosa Pine 2.4 Bottomland flats, Playas, Saline Flats, Alkaline Big Sagebrush Flats 2.5 Mesic and Dry-Mesic Bottomland Needlegrass-Wheatgrass Prairie with Occasional Cottonwoods 2.6 Prairie Dog Colonies (not mapped)

GROUP 3. PONDEROSA PINE 3.1 Ponderosa Pine Savanna 3.2 Ponderosa Pine-Juniper-Dry Prairie Mosaic with dominants depending on location in the landscape 3.3 Ponderosa Pine Slopes with Prairie Meadows and Occasional Douglas Fir Ravines

GROUP 4. MISSOURI RIVER AND TRIBUTARY STREAMS: SLOPE TOES, COLLUVIUM AND DRY TERRACES (above floodplain) 4.1 Colluvium and Toe Slope Shrublands 4.2 High River and Stream Terrace Shrubland and Prairie

GROUP 5. SALINE DRY DRAINS 5.1 Saline Dry Drains with seasonal flooding events

GROUP 6. SMALL STREAM AND UPLAND WETLANDS 6.1 Intermittently Flooded Lakes, Marshes and Vernal Pools 6.2 Wooded Wetland Ravines and Drains 6.3 Small Stream Wetland Mosaic Structured by Beaver and Fire

GROUP 7. MISSOURI RIVER AND TRIBUTARY STREAMS FLOODPLAIN COMPLEX 7.1 Low Terrace Cottonwood Flats 7.2 River and Small Stream Wetlands (moist soils)

Changes in vegetation since European settlement. The most important changes from historical types include 1) the loss of beaver-maintained wetlands in 6.3, 2) a likely decrease in patches kept open by prairie dogs in 2.6, 3) an increase in woody stem density and cover related to the early 20th century reduction in fire frequency in 2.3, 3.1, 3.2 and 3.3 and 4) clearing of bottomland cottonwood flats for making hay in 7.1. And there are fifth, the changes in species composition and cover in the grasslands types brought about with introduction of exotic plant species and changes associated with the increase in grazing intensity by introduced livestock.

PRE-EUROPEAN FIRE REGIMES OF CHARLES M RUSSELL NWR There are many ways to get at original fire frequency. The ones available or with most useful application vary from site to site. Table 2 shows the kinds of evidence that proved useful for reconstructing past fire regimes at CMR. How these kinds of evidence were used to put together the map of presettlement fire frequency is covered in the section on mapping below.

Table 2. NINE KINDS OF FIRE HISTORY EVIDENCE AVAILABLE AT CMR 1. The three fire-scar chronologies 2. Area lightning ignition patterns 3. 27 years of refuge records of ignitions and area burned 4. Fire frequency indicator plant communities such as those with Juniperus horizontalis or Douglas fir 5. Likely fire frequency indicator species such as Echinacea angustifolia and Orthocarpus luteus 6. 1/10 hectare study plots 7. Tree demography studies 8. Historical photos and matching photos taken later 9. Historical descriptions by Lewis & Clark 1805-1806, Larpenteur 1833-1872, Schultz 1877-1901 and others

Effects of the 2006 Wildfire at Soda Creek in Old Growth Douglas Fir and Ponderosa Pine

In June, 2006 I completed a 1/10 hectare study plot on the south side of Soda Creek near its mouth. The plot was located on the upper north slope and included a ravine. This was in the first weeks of the study and I explained to my wife Vonda that this stand, on a cool, north-facing slope, with its old trees, protected by the Missouri River to the north and east and somewhat by the Soda Creek bottom, was likely one of the more fire sheltered sites at CMR. Three weeks later it burned down. Figures 1 through 4 show before and after photos of the site, the first pair looking toward the northwest and the second toward the northeast.

Figure 1. Western end of study plot CMR04 at Soda Creek one year after the July 15, 2006 wildfire. It was a wind-driven, cold front fire, generating intense heat as it came upslope from below. There was 100 percent mortality: not a single woody stem in the plot survived. This is an exact match to the figure below.

Figure 2. The same plot on the day of data collection three weeks before the fire. All plant species present were identified and their cover values recorded by layer. The ravine was dominated by Douglas fir with ponderosa pine on the side slopes. Note that everything on the ground was consumed completely, including the large trunk in the foreground. Two

11 years later, in July, 2008, all stems were cut and sections collected in a demography plot 5 m x 50 m, representing half of the original study plot.

Figure 3. The eastern end of the same study plot at Soda Creek after the fire. The photo below is an exact match from before the fire.

Figure 4. Eastern end of 1/10 hectare plot CMR04. Many decades without fire have permitted accumulation of a lethal, continuous cover of Rocky Mountain juniper and saplings of fir and pine in the shrub layer. Taller midstory saplings of both tree species added additional intensity when fire returned. Grasses and forbs are almost absent, having

12 been shaded out in the long absence of fire. Note also in the matching photo above the complete consumption of all living shrubs, including their live stems, leaving no aboveground trace of their existence after the fire. The dead wood on the pine “elbow” at upper left caused the whole live stem to be burned off and it can be seen on the ground in the photo above.

Figure 5. Location of the figures above, taken before the fire, looking across the Soda Creek valley to the study site in a partially fire sheltered cove (arrow) on the heavily wooded north-facing slope above mouth of Soda Creek. A portion of U.L. Bend can be seen on the far left.

While visiting the Soda Creek fire site to collect sections of fire–killed trees to prepare a fire scar chronology, Mike Stambaugh of the Missouri Tree Ring Laboratory asked me in reference to the whole 121,000 acre fire complex and, in particular, the severity of the fires in wooded areas, “Is this natural?”. The question gave me considerable pause because there were several possibilities. Frequent fires are cool fires because there is not much time to accumulate more than light grass, pine needle litter and a few small shrubs to burn. Longer fire intervals lead to more intense fires as shrubs and other woody fuels have time to form increasingly deep, flammable layers beneath the trees. One explanation, if portions of CMR were such areas, might be that the event merely represented the natural 100 year fire. Or was it that the intensity was the result of alterations in the natural fire regime caused by livestock grazing and fire suppression? Or might it be an effect of global warming with higher summertime temperatures? I could think of arguments to support each of those possibilities. The rest of this analysis will attempt to answer Mike’s question.

2) THE PHYSICAL FIRE ENVIRONMENT AT CMR

The fire landscape is far more complex than just its topography. To condense the factors into a simple equation, Fire = f(R*V*T*A*I*C*W*L), where fire is some (undefined) function of Relief (topography), Vegetation, Time, Animals, Ignition, Climate, Weather and Landscape. Fire topography includes relief, the distance from the lowest local part of the landscape to the highest and the steepness of elevational changes which can affect whether fire is likely to spread. Ways to examine this spatially include mapping the Topographic Roughness Index (see Figure 20 below) which is something like a slope map that plots the difference between local highs and lows. Another is a land surface form map, developed by a geographer, Hammond (1964) for which any point is the sum of the distance from the lows to the highs, the percent of land that is flat or only gently sloping, and whether the flat or gently sloping part is on the uplands or in bottoms. It wasn’t Hammond’s intention, but it turns out that these characters can be used to define portions of a landscape with fairly uniform fire behavior and this was use to make a map of presettlement

13 fire regimes of the US (Frost 1998). This is only a very broad brush technique and much more detail is needed for fine-scale mapping s at CMR. Vegetation includes production of dead and live fuels and the structure of living vegetation: one-layered vegetation such as grass or shrubs, or multi-storied woody vegetation produce very different fire intensities and behavior. Time relates to the time for fuel accumulation, whether the minimum time for enough fuel to be produced to carry fire at all, or the time that allows accumulation of fuel for high intensity fires. Consideration of Animals ranges from soil microorganisms that accomplish fuel breakdown to grazers such as bison that eat enough potential grass fuel to reduce frequency. “I” refers to ignition frequency and ignition sources such as native peoples versus lightning, and ignition generators such as mountains. Climate covers items such as the length of the fire season and thins that affect fuel ignition probability like temperature, rainfall and humidity. Weather is a subset of climate, that looks at the things that affect a particular fire such as temperature, windspeed, wind direction, and factors affecting fuel moisture and fire behavior. L for landscape summarizes emergent properties such as fire compartment size, which partially controls fire frequency, as well as firebreaks, fire filters and pathways for fire flow.

Looking at all the interacting factors that produce fire frequency at any point on CMR is an exercise in reading the fire landscape. We’ll start with looking at firebreaks and fire filters.

FIREBREAKS AND FIRE FILTERS • Firebreaks – The Missouri River – Beaver wetlands – Barren clay/shale and rock outcrops • Fire Filters – Soil mapping units containing Badlands or Rock Outcrop as the first or second most abundant component – Soils with productivity too low to reliably carry fire.

In the current jargon of ecology, fire at CMR is controlled by both bottom-up and top-down factors. The bottom-up forces are represented by geology, which gives rise to soils which support either lush grasses that promote fire spread, leading to high fire frequency, or soils so poor and sparsely vegetated that they act as firebreaks or fire filters. Bottom-up drivers for any spot are the local slope, aspect and fuel type and fuel continuity, whether of pine needles or grass, which is a product of local soil productivity for grass fuels, the connectivity of which leads to fire spread and higher fire frequency or, conversely, the amount of bare ground which has the opposite effect. This category also includes the abundance of fire filters such as prairie dog towns, gray clay gumbo soils or badland soils upwind of the site in the pathway of fire approach reduce fire frequency, also the length of fire paths, fire compartment size and orientation of the compartment to prevailing winds during fire season.

Top-down forces are represented by lightning and human intervention. Top-down drivers of fire frequency at CMR are the four mountain groups that serve as lightning generators, length of fire season and climate: July temperatures high enough and precipitation low enough to cure live grass fuels and dry out woody fuels

Geologic Features Related to Firebreaks and Fire Filters The major geologic formations of CMR include, from bottom (oldest) to top: 1) The Bearpaw Shale is marine in origin, Upper Cretaceous in age and around 1100 feet thick (?), the upper member is 420 feet thick and is sandy in some of its uppermost parts that appear at the eastern end of CMR. This is the dark gray shale that dominates the eastern half of CMR and underlies the younger formations in the eastern half. Since the formations dip to the southeast, the Bearpaw shale eventually dips under the more light colored formations that occupy the surface in McCone County and parts of Garfield.

Figure 6. Nearly unvegetated Bear Paw Shale form steep lobes along the north side of the Missouri River. The barren walls just above the flat river terraces serve as local firebreaks partially preventing spread of fire from one section of bottomland to another. Deep ravines provide fire-refugial habitat for Doulas fir. Seen from the eastern end of Knox Ridge, west of US 191 bridge.

2) Fox Hills Sandstone, Upper Cretaceous, shallow marine, a relatively thin layer at CMR, it crops out along the Missouri River and Big Dry Creek. The upper part, 80 ft thick, is the bed of soft brown sandstone that contains the massive round and tubular concretions up to three feet in diameter conspicuous along one section of highway 24 south of Fort Peck. This stratum also contains coal beds and dinosaur remains

3) Lance Formation – shale and sandstone of non-marine (lacustrine, fluvial & swamp) origin, 900 ft. It is lower part is the Hell Creek Member.

4) Fort Union Formation, 700 feet thick, contains thin layers of dark shale with alternating thick layers of white sandy clay, sandstone and coal seams. This surface was uplifted, eroded and then mantled in places with younger sediments of Oligocene to recent age. It extends to Theodore Roosevelt National Park and is the dominant formation there and was named for Fort Union, the American Fur Company headquarters on the Montana-North Dakota line where it also crops out. The younger formations have light and dark colored layers that are soft and differentially erodable, leading to the highly dissected landscapes of the badlands. Badland is most common in semiarid and arid regions where streams are entrenched in soft geologic material. Local relief generally ranges from 25-500 feet. Runoff potential is very high and geologic erosion is active. For more detailed geologic information see the see McCone soil survey, p. 211- 215.

15 FIREBREAKS

Figure 7. A landscape-scale firebreak. Gumbo ridges in Valley County prevent fire that might otherwise spread into the refuge from lands to the north and west. At least eight barren clay ridges can be seen receding into the distance, each acting as a firebreak. Along road from Willow Creek to Fifth Ridge.

FIRE FILTERS

Figure 8. Fire filter soils at the point between Hell Creek (left side) and Crooked Creek (right). The darker shade of gray represents soil units with a high percent of rock outcrop and barrens with too little fuel to reliably carry fire. The lighter color

16 represents light-colored badlands soils with similar characteristics limiting fire spread and therefore also reducing fire frequency. Such soils are much more common in the eastern third of CMR, contributing, along with a lower lightning ignition rate, to much lower fire frequency at that end.

Figure 9. Badlands on the north side of Haxby Road on the way to Haxby Point.

In Figure 9 few fires would make their way through this fire filter. A local ignition by lightning might burn a small patch. Fires spreading in from the southwest (see grass and small sagebrush to lower left), the prevailing direction of fire movement during fire season, might find some small channel to make it past the barren ridge or it might go out. Another fire might spot over or find a way around but be stopped by the incised ravine in the middle distance. Yet another fire might make it past those two barriers but be stopped by the high ridge in the distance. The net result is that not all fires would make it through this fire filter, with the result that the lands downwind would experience a lower fire frequency than if there were continuous fuel.

Fire Frequency Versus Fire Severity

Fire frequency is determined by ignition rate, fire compartment size, fuel continuity and the presence or absence of firebreaks and fire filters. Fire severity is affected by wind, temperature, fuel moisture, fue arrangement and the amount of fuel On the prairies, the higher the wind, the faster the rate of spread on soils such as loams and fine sandy loams that are more productive of dense, continuous grass fuels and the higher the likelihood that fire will carry through fire filter areas—zones of low productivity, prairie dog towns and prairie badlands. Wind little affects intensity in shortgrass prairies but can increase intensity greatly when fire reaches timbered and shrubby communities. Fire frequency is very complex in shortgrass prairie regions and is studied in detail below

Given that nearly all fires of any size at CMR occur in the narrow window of July and August, when conditions are often very hot and very dry, variation in fire severity follows three main factors: wind velocity, wind direction and time since last fire, more time allowing more accumulation of fuel, at least in the woody communities. Fires following dry cold front ignitions are more intense on average than others

17 because winds are strong and sustained. Further, cold front fires are pushed into areas such as northwest and north slopes and ravines that are missed or only lightly affected by southwest fires when wind velocity drops after small storm passage. The result of lower fire frequency on these facets of the landscape is that intensity is higher when a fire driven by northwest or north winds finally does reach them.

The most severe fires of all occur in the pine breaks when there is a fire interval of three to six decades or more with no fire, providing time for accumulation of juniper and Douglas fir saplings in the shade of ponderosa pines, setting the stage for stand-destroying fire as seen in the paired photos of the fire at Soda Creek. It would have been natural for this to occur in a small pocket missed by fire here and there but not for the broad outbreaks of stand-destroying fires that occurred in July 200.

This discussion applies in particular to the higher, drier parts of the breaks landscape. There are lower, cooler and moister pockets in deep ravines along the Missouri River and the lower ends of its deeply incised tributaries where fire may have no access for hundreds of years. The result is a continuum of fire frequency and severity from the prairies and upper breaks (frequent, light fires) and the most fire-sheltered ravines (rare, severe fires).

Stand-destroying fires are a natural component in the complex pattern at CMR. Consider the extreme continuum of sites available at CMR, some very fire exposed, some sheltered from fire in deep ravines. On the fire frequency map there were 134 fire compartments. The boundaries of fire compartments are not absolute, they merely serve to demonstrate relative fire frequency and no particular fire is necessarily likely to respect the boundary. Also, given the variation in topography there may be many facets of variation in fire frequency and intensity based on the different slopes and aspects found in a compartment. If we looked at each variant of slope, aspect, fuel continuity and degree of fire exposure and fire shelter we might have on the order of 15,000 fire behavior units at CMR. Thinking of these as pixels, any particular fire, with its own set of weather and fuel conditions would affect only some local groups of pixels and might affect individual but similar pixels, such as steep northeast-facing slopes differently.

While small stand-destroying fires can occur naturally at CMR, under the historical fire regime they occurred in somewhat random small units, a particular slope or ravine that by chance had been missed by the last several fires. When fire does eventually reach that pocket it could burn out completely, killing everything including trees 200 years old, but this would not happen everywhere at once because in much of the landscape recent fires would have kept the understory cleaned out of woody fuels.

Similarly, while the deep, fire sheltered ravines along the Missouri River make small lightning targets, rare ignitions from lightning strikes can occur anywhere so the security of such sites in not deterministic and once in a while one of these will burn out. The resulting pattern is that the pockets of Douglas fir have different ages, some having survived with the luck of the draw to be 300 or more years old.

The most critical factor contributing to intense fires is the interval since last fire and as we shall see, the fire complex of July 15, 2006, in combination with sustained strong winds that persisted for days, encountered un-natural accumulations of fuel.

IGNITION SOURCES

Native Americans as an ignition source at CMR

There is little in the historical records I surveyed to suggest any effect of First Nations peoples on natural fire regimes in the CMR region other than their indirect effects through reduction in number of bison. With fewer bison there should have been more grass and the more grass there was, the better the fuel connectivity and that should have led to higher fire frequency.

18 In all of James Schultz’s writings during his years of living with the Blackfeet, there is no mention of their use of fire other than for heating and cooking in the village. Nor is there discussion of fire in John C. Ewers book “The Blackfeet”. In the only one mention of fire in the CMR area by Lewis & Clark, a burning tree almost fell on a tent in river bottom. The cause of fire, whether wildfire or escape from a campfire was not explained. One painting by Charles Russell depicts Blackfeet setting fire to the Crow buffalo range but I read nothing to indicate that this was a common practice. War parties were extremely careful with fire, building circular “war houses” of sticks and logs in the pine breaks of the Missouri River to shield their small cooking fires form detection by the enemy. The journals of early trappers and traders give no instance of Blackfeet using fire to drive buffalo even though the fire surround had been common on the plains to the east before the advent of the horse. Instead the Blackfeet chased buffalo, originally on foot, and then after 1750 on horseback, into a V-shaped funnel with side walls composed of rock and brush. Once in the funnel Indians on both sides rose up waving blankets and hides to excite the buffalo into a stampede toward the narrow end where they plunged over a precipice where any that survived the fall were killed. Sometime around 1820 the Blackfeet abandoned this method in favor of killing buffalo individually from horseback (Schultz 1962).

In determining original fire frequency at CMR we have to consider two cases, first is the use or non-use of fire in the 150 year pre-European period, 1730-1882 after acquisition of the horse. Second, what use of fire did the Blackfeet, Crow and their predecessors make in the much longer (by at least 6000 years!), pre-horse era?

Nomadic existence and fixed-village agriculture led to very different uses of fire. Downstream, the Mandans, before they obtained horses, annually burned the prairie near their permanent villages to green up grass to attract buffalo to the vicinity of the villages for hunting (((H__get the reference summarizing the historical accounts))). Before the advent of horses they also practiced the use of the fire surround to kill buffalo with bows and arrows in a confined space within a ring of fire. Other Indians of the Willamette Valley in Oregon burned every year as part of their annual fall gathering of food for winter. After the fire passed, acorns were exposed and easily gathered by the bushel for storage, and fire burned off the toxic coating of seeds of tarweed (Madia spp.) making it easy to gather another major food source.

In contrast, fire was a liability to nomadic buffalo hunters such as the Blackfeet and Crow. Once winter camp had been set up, any late fall wildfire would have been a disaster because it would consume the forage the horses needed to survive the winter. Landscape fire would have been of no use in their fixed winter camps where, because of snow and the grazing of hundreds of horses, there was nothing to burn anyway, and by the time of spring greenup the village was ready to pack up and move. By then any buffalo in the area had typically been driven away by winter hunting and the bands were ready to go to summer hunting quarters near the buffalo. Even so, after the snow melted, they remained in their winter encampments until the spring grass greened up and their horses had time to graze and recover from the near starvation of winter. Only after the horses had fattened did the entire village pack up and move to summer buffalo hunting range. Once there any accidental fire could burn off the grass causing the buffalo to move away to where grass was still available. This would require dismantling the entire village and moving to a new summer hunt location, a large undertaking. One year, large prairie fires in the vicinity of the Cypress Hills to the north of CMR in Saskatchewan caused the buffalo to move south, ruining the trade in buffalo robes in the area that season.

If neither the Blackfeet nor Crow used fire outside their villages, that leaves lightning as the sole ignition source in the region. This simplifies the interpretation of natural fire frequency which is already complicated enough. The role of topography, lightning and other factors is discussed below.

19 EFFECTS OF TOPOGRAPHY ON FIRE FREQUENCY

Table 3 shows local relief at points around CMR. The eastern half has generally less than half as much relief as the western. Relief ranges from less than 180 feet toward the eastern end of the refuge to its extreme around 1027 feet, or more than 5 times as much, in the badlands east of Seven Blackfoot Creek. Another area of high relief includes the deep river gorge and ravines upstream from highway 191 to Grand Island. The Soda Creek area downstream has about 1/3rd less relief.

The Willow Creek bottomland north and east from Fifth Ridge is some 22 meters lower than the original lake level. The remains of experimental berms from formerly irrigated lands can be seen along Willow Creek Road at a point just north of Duck Creek Bay, left over from the period when water could just be gravity fed down from the high level of the lake. These plots are exceedingly barren now, perhaps the result of salinization from years of irrigation and from the Vaeda clay substrate which has a very low productivity for vegetation.

Relief has a strong effect on fire frequency through its effects on fire spread. In general, flat to gently rolling landscapes have the highest fire frequency as long as fuels such as prairie grasses are continuous. Fire doesn’t move downslope well and with increasing topography in the CMR area there are more areas of rock outcrops and zones of clay too barren to support grassy fuel.

LOCAL RELIEF (in meters and feet):

Location High point minus Relief low point (feet) Upper Missouri River Gorge 900-700= 200 m 656 near US 191 Soda Creek 830-685 = 145 476 Musselshell River near 830-685 = 145 476 southern refuge boundary Mickey Butte 882-685 = 197 646 Lost Creek 870-685 = 185 607 East of Seven Blackfoot Creek 998 - 685 = 313 1027 Hell Creek 820-685 = 135 443 Middle Eighth Ridge 800-685 = 115 377 Fifth Ridge 740-685 = 55 180 Near mouth of Big Dry Creek 760-685 = 75 246 Nelson Creek 760-685 = 75 246

Table 3. Amount of local relief for various points around CMR, shown as the difference from the highest point to the level of Fort Peck Lake, or in the upper parts, the level of the Missouri River.

Figure 10 below is a model of topographic roughness for CMR. The Topographic Roughness Index (TRI), developed by Richard Guyette and Mike Stambaugh of the Missouri Tree Ring Laboratory was applied to the topography of CMR (Stambaugh and Guyette 2008, Guyette and Dey 2000).

Figure 10. Topographic roughness model for CMR. Red areas represent high relief, with greater distance from the high points to the low points, and with steeper slopes. Blue is low relief with the darkest blue of the river and lake representing perfectly flat. Data is missing for UL Bend and Hawley Flat but that area, with its low relief would be mostly blue.

Using just the element of topography we would expect the highest fire frequencies to be in the blue areas and the lowest in the red areas. The two regions of lowest fire frequency should be found at the western end of the refuge in the red areas of the river gorge upriver from the Sand Creek section to Grand Island, and again in the red, high-relief areas of the central portion of the refuge centered around the mouth of Seven Blackfoot Creek. The most fire-frequent areas should be the blues in Valley and McCone counties. If the refuge extended further north and south, the upland prairies in all the surrounding landscape would be blue for many miles, representing their flat to gently rolling topography, leading to the expectation that they should have the highest fire frequency of all.

When, however, we examine all factors, including 1) zones of poor fuel connectivity; 2) the relation of any point to the major ignition sources, and 3) fire compartment size and orientation to prevailing winds, we see that topographic roughness as a predictor of fire frequency is often overridden by other factors. It seems that nothing is ever simple in shortgrass prairie.

In comparison with the CMR fire frequency map (GIS map 2), the expected relation of the model to fire frequency holds true in several places. These include the deeply fire sheltered ravines with Douglas fir at the western end and the frequent fire area prairie areas to the south of Sand Creek and along the western side of the Musselshell River and in the Lost Creek area just south of the more rugged lands along the Missouri River. The red areas, especially those immediately paralleling the river and its tributary creeks

21 and coulees define the zones, within which, in the deepest and most fire-sheltered pockets and ravines, are found the habitats for Douglas fir.

The expected relation of the model to fire frequency, however, does not work in several conspicuous areas. despite its blue color, Valley county from Middle Eighth ridge nearly to Fort Peck is essentially fire-free. This is because almost this whole quadrant of CMR on the north side of the river is almost uniformly underlain by what seems to be a particularly sterile member of the Bearpaw Shale. In may places the only plants are tiny halophytes such as Atriplex dioica which contribute nothing as fuel (Figure 11). So, despite the gentle topography, the fuel matrix is reduced to pockets and patches of grass with insufficient continuity to permit fire spread no matter how dry and windy condition may be. Frequent ignitions may occur but can only burn the local fuel patch ranging in size from a few square meters to a few hectares.

Figure 11. Study plot on a clay soil in Valley County dominated by Atriplex dioica, a small spring annual here largely dried up and giving a reddish cast to the sparse vegetation layer.

An example can be seen in the fire on Middle Eighth ridge in the July 15, 2006 lightning fire complex. Middle Eighth is on the western border of this infrequent fire area where fuels are beginning to coalesce into larger patches. Figure 12 shows the easternmost ignition in that fire complex (the tiny red spot seen in Figure 20 below). The lightning ignition occurred in a stand of ponderosa pine and burned to the southeast (right side of photo). It was unable to back across a small gully on the left side and covered only a few hectares in any direction before encountering sparsely vegetated gray clay and going out despite the accompanying strong winds.

Figure 12. Lightning fire at Middle Eighth Ridge

LIGHTNING AS AN IGNITION SOURCE AT CMR

In a typical year, over the past 27 years for which records are available, between 2000 and 3000 acres have been burned in wildfires. This is of course with fire suppression. Without suppression we might expect at least twice that amount to burn or 4000-6000 acres (Paul Pallas, pers. comm.). Furthermore, there are fire pathways leading into CMR from the south, which, in the original landscape would have been unbroken by roads and overgrazed landscape, leading to perhaps another doubling to 8000-12,000 acres. Averaging in the occasional wind-driven cold front fire that covered more territory, might bring the average amount of land burned annually in CMR in pre-European conditions to at least 16,000 acres. This would give a fire return interval for the whole refuge of about 69 years. However, fire is very unevenly distributed at CMR. Some of the deep ravines with Douglas fir, where there are trees several hundred years old had much longer intervals and other areas such as the badlands were essentially fire free but for small spot fires had nowhere to go. The majority of fire were focused in the more fire frequent areas shown as fire frequency classes A, B and C in the GIS map of fire. We’ll look at the causes for this heterogeneity of the fire pattern.

Ignition Sources and Fire Movement

There is a certain background of random ignition from summer convection storms, as can be seen on the prairies to the east of Fort Peck. These may come most often on south winds when there is sufficient moisture in the air and enough mid-day heat to generate small storms drifting with the winds. Movement would be mostly from south to north, and some such fires could be carried by prairie grasses into the half of the refuge on the south side of the Missouri River while others are ignited within the refuge. If this were the only source we would expect to see a fairly regular pattern of fires but this is not the case.

In the CMR region two dominant fire tracks could be expected, one along an axis of post-cold front fires generated by lightning from the Little Rocky Mountains, moving from northwest to southeast and the second along an axis of fires originating downwind from the Snowy Mountains and the Judiths, moving

23 from southwest to northeast. On typical fire season days of July and August, winds from the southwestern quadrant often contain enough moisture to generate small pepperpot convection storms but not widespread rain. These small cells generate lightning and the resulting fires in the ignition fan downwind from the mountains from southwest to northeast until reaching the Missouri River. Few of these fires would be expected to spot across the river because the winds associated with such tiny storms, while locally intense, dissipate immediately on passage of the storm cell, leaving a return to the general regional air flow and sometimes even a dead calm. Summer orographic and local convection storms should create their highest fire frequency just in the western half of the refuge, including the Sand Creek area, and on either side of the mouth of the Musselshell River in the center of the range of influence of the Judiths and Snowies.

On the northern, downwind side of the refuge, ignitions from the southwestern source have a single component, that of ignitions that occur within the boundaries of the refuge on the north side of the river. This is because few fires ignited farther north from the refuge would be likely to back that far south. On the southern, upwind side of the refuge, natural fire frequency would consist of the additive effects of such local ignitions plus fires originating at various points upwind between the refuge and the mountains, filtering across the prairies and spreading into the refuge. So for the eastern half of the refuge we should expect a little higher fire frequency on the south side of the river than on the north.

In the second track, cold front fires can be ignited on southwest, west or west-northwest winds. Since cold front winds swing rapidly on frontal passage from southwest to west and northwest, fires ignited immediately downwind from the Snowies and Judiths would curve to the east and then southeast, never reaching the CMR.

While orographically spawned storms from the Snowies and Judiths on normal southwest winds would only rarely be likely to spot across the Missouri River, the likelihood increases greatly with cold front fires. The lower peninsula of U.L. Bend, which burned in the July 15, 2006 fires may have been ignited by spotting across the river from the burning timbered ridges above Soda Creek. With the strong post-cold front winds experienced that night and lofting of small firebrands such as glowing flakes of bark, long- range spotting becomes much more likely.

Lightning Ignition Fans

The Little Rocky Mountains come into play when cold front lightning activity occurs on west and west- northwest winds. In the first case fires ignited downwind from the mountains may curve south into the refuge as post-cold front winds shift to the northwest and north. This effect should almost entirely miss the western Sand Creek section because it is almost due south of the mountains, whereas cold front ignitions, occur to their east and would initially travel to the east and then southeast. The ignition phase would be long over by the time winds shifted completely to the north. It is likely, however, that a few fires crossing the prairies after cold front ignitions downwind from the Bear Paw Mountains would have reached this area.

Figure 13. The Little Rocky Mountains generating lightning on southwest winds on a tranquil August day. Six phases of orographic lightning generation can be seen. A) moisture laden air with only minor clouds formation of “fair weather cumulus” formed by local convection updrafts can be seen, B) a cumulus puff builds over the first uplift zone on the southwestern end of the mountains, C) this builds within only about 20 minutes to a much larger cumulus which, as it begins to flatten out (D), generates its first lightning strokes and begins to drift away from the mountains downwind (E) producing lightning and a small amount of rain. This continues for a distance of 50-100 miles, after which the thunderstorm begins to collapse and lightning ceases (F).

There are four prominent sources of lightning activity within range of CMR and these serve to raise the frequency of lightning ignitions within their ranges of influence:

Source 1. The nearest lightning source in the area is created by the Little Rocky Mountains (Figure 14). With dry cold fronts, fires originating on the eastern downwind flanks of the Little Rocky Mountains and in the downwind landscape with fires can be carried by prairie grasses into the center of the refuge on the north side and ignitions occur within the refuge, as on the evening of July 15, 2006. The Little Rocky fire fan would be expected to influence primarily the center of the refuge, with some effect on the eastern half but little in the west.

Figure 14. The Little Rockies project an ignition fan that extends further east than the other three below.

Figure 15. The resorption phase of an orographic storm cloud. While lightning- generating clouds can be formed out of clear air by uplift on the upwind side of a mountain they can also be resorbed, leaving no trace on the downwind side. A typical range limit for downwind lightning is about 100 miles and it may take another 100 miles for the cloud to dissipate completely. I followed this cloud downwind from its source over the Black Hills of South Dakota, where there was active lightning and some rain for the first 50-75 miles, while the tip, dwindling into the distance above as its moisture was taken back into the dry prairie air, ended about 200 miles northeast of the mountains.

26 When there is only enough moisture in the air for the mountains to generate small lightning cells in summer or during passage of dry cold fronts, lightning generation typically dies out within about 100 miles downwind. So the western two-thirds of the refuge is within range of the mountain effects. Where prairie grasses to the south are continuous, fire frequency is augmented by those fires initiated upwind that are able to filter into the area. Natural firebreaks and fire filters posed by complexes of partially barren badland soils cut off the potential for this kind of fire flow in many areas in Valley and Garfield Counties and on the southeastern fringe of the McCone portion of the refuge.

Source 2). The Judith Mountains to the south comprise the second source and with the Snowies, form one of two largely ovelapping ignition fans, both originating with orographic lifting of moisture-containing air by isolated mountain groups. The Judiths are closer and the Snowies a little farther away but higher and both can be seen generating lightning and thunderstorms on southwest winds on any summer day when there is enough moisture in the air.

Figure 16. The sphere of influence of ignitions originating from the Judiths overlaps that of the Little Rockies, influencing primarily the western half of the refuge. It fills in the ignition gap to the south of the Little Rocky Mountains.

Figure 17. The pattern from the Snowy Mountains overlaps much of that from the Judiths and a small portion from the Little Rockies in the center of the refuge.

Source 3 is a third ignition fan to the southwest originating from the Snowy Mountains. It largely augments the Judith Mountains fan except in the area around the mouth of Seven Blackfoot Creek. As mentioned above, these two southwestern sources produce fire in the refuge in two ways. First by ignitions within the refuge, on both sides of the Missouri River, and second from fires originating in and downwind from the Snowy Mountains and the Judiths, carried by prairie grasses into the southern half of the refuge on southwest winds. The result should be a higher fire frequency on the south side of the Missouri River than on the north. The associated winds should produce mostly fire movement from southwest to northeast, with fires only rarely crossing the Missouri River. The few fires ignited within the refuge on the north side or spotting across the Missouri River would be carried out of the refuge onto the prairies to the northeast. The fourth ignition source is the Bearpaw Mountains cluster (Figure 18 below).

Figure 18. With all four fans superimposed, what stands out is the white area in center where all four overlap. This should be the ignition hotspot for the refuge. A much lower rate would be expected for the Hell Creek area which falls within the sphere of influence only of the Little Rocky Mountains and that is a cold front pattern, only functioning on dry cold fronts and produced fewer fires than those on the prevailing winds from the southwest. .

The behavior of fires under all conditions above assumes just enough atmospheric moisture for lightning with limited rainfall downwind from the ignition source. Obviously on very rainy days fires are not likely to propagate.

Figure 19. Lightning strike density of the U.S. CMR is in the general region of the northwest that receives 0.5 to 1 strike per square kilometer per year (dark green) but in the CMR area, there are areas with 1-2 strikes per square kilometer per year (light green). For the approximately 1.1 million acres (4452 km2) of CMR that would 4452 to 8904 lightning strikes per year, of which only a very small percent would produce ignitions.

27 Year Fire History 1980-2007 – the Actual Distribution of Ignitions

Figure 20. Fires in the eastern third of CMR 1980-2007. The large fire (red) on the south side of Sutherland Creek Bay, the red area to the southeast across the river and the tiny red dot on Middle Eighth Ridge on the north side of the Missouri River were all ignitions in the July 15, 2006 fire complex resulting from lightning generated by the Little Rocky Mountains during passage of a dry cold front. Comparison with Figure 18 above shows that all three fires lie close to the furthest extent of the lightning ignition fan originating from that source. To the east, the only ignition source is lightning associated from random small, summer convection storms. These are the main source of ignition across the prairies from CMR, east to the Mississippi River. Fires in the eastern third of CMR also are limited in size because the presence of interfingering badland soils limits fire spread.

Figure 21. Fires in the western third of CMR 1980-2007. Fires are more numerous here because the western third lies within the overlap area of three ignition fans, those of the Judith Mountains, the Snowy Mountains and the Bearpaw Mountains. The little Rocky Mountains, although the closest lightning generator, has no influence west of the Fisher Coulee and Nichols Coulee area because lightning fires occurring on the prevailing southwest winds only occur on the north and northeast sides of the Little Rockies and cold front fires would occur on the east sides. By the time cold front winds had swung around to the north, as would be needed to affect this area, the lightning phase would be long past. The largest fire, on the east side of map, looks like a post cold front ignition from either the Bearpaw or Little Rocky Mountains that started at the top and then swung around with shifting cold front winds that drove it south to the river. Disregarding human-caused fires, the others may all be prevailing winds fires from lightning originating from the Snowies and the Judiths. These small storms should be expected to burn less area than cold front storms because, while there may be strong winds associated with the storm, as soon as the storm cell passes on the winds die down to calm or the normal background breeze. Fires with little wind tend toward more circular and broad elliptic shapes than cold front fires and are more easily suppressed.

Figure 22. Fires in the middle third of CMR 1980-2007. This is the hot spot of the refuge, being the only portion within the overlap of all four ignition sources. Most of the fires shown in red originated in the July 15, 2006 complex. Strong winds persisted over several days following frontal passage, resulting in large areas burned. They show a pattern of movement from northwest to southeast, complicated by rotation as the winds shifted to the north and then northeast over several days. The large Soda Creek Fire, ignited on the same night, on the west side of the Musselshell River is missing from this illustration.

Implications for fire elsewhere in shortgrass prairie. Given these four additional ignition sources, fire frequency at CMR should have been higher than in SGP to the east. This would only be true, however, where there were similar patterns of soils with productive, lush grass soils interspersed with barren soils to serve as fire filters or firebreaks. In more fertile portions or in areas transitional to midgrass and tall grass prairie in the Dakotas fire frequency could be higher even with a lower ignition rate because of the larger fire compartment size. A fire compartment is defined as a unit of the landscape with no internal firebreaks and fairly continuous fuel so that an ignition in one part might burn the whole thing unless put out by rain or high nighttime humidities. It is a rule that the larger the fire compartment the higher the fire frequency. In contrast, in landscapes divided into many small compartments by streams, gullies, badlands or zones of soil too poor to reliably carry fire (such as in parts of central and northeastern Garfield County), it takes many more ignitions to burn the same amount of land.

32 3) FIRE RELATIONS OF KEYSTONE NATIVE ANIMAL SPECIES: Beaver, Prairie Dogs, Bison

There was clearly interaction between native grazers and fire in the shortgrass prairie of the CMR area. Of these, the most important seems to have been a natural suppression of fire frequency resulting from bison and other native grazers such as pronghorns, deer and elk consumption of the grass fuels that carry fire. Other factors are a likely slight increase in fire frequency after collapse of beaver wetlands that served as local firebreaks and a reported late 19th century increase in prairie dog numbers after poisoning of their predators which could have reduced fire frequency as a result of more fuel free colonies to act as fire filters.

Some Animal-Fuel-Fire Cascades

1) Wolves ▲ Coyotes ▼ Prairie dog towns (fire filters) ▲ Fire frequency ▼ 2) Wolves ▲ Bison ▼ Grass height and fuel continuity ▲ Fire frequency ▲ 3) Wolf trapping and poisoning ▲ (1830’s1930, initially for pelts, then for intentional extirpation) Wolves ▼ Coyotes ▼ other poisoned predators of prairie dogs (mountain lion, bobcat, badger, hawks, eagles) ▼ Prairie dogs ▲▲ Fire frequency ▼ 4) First Nations peoples’ use of bison ▲ grass height and fuel continuity ▲ Fire frequency ▲ 5) Mild winters  Bison ▼ Grass height and continuity ▲ Fire frequency ▲ (next summer) Coincidence of a previous mild winter followed by a hot, dry summerProbability of a fire year ▲ 6) Trapping of beaver ▲ firebreak wetlands deteriorate & collapseFire compartment sizes ▲ Fire frequency ▲ (probably minor) 7) Spread of Spanish horses among Blackfeet (1730-1750)(last buffalo “pounds” or “jumps” used by the Blackfeet in 1830’s)  Bison ▼ grass height and fuel continuity ▲ Fire frequency ▲ (did hunting from horses have any more or less effect on Bison populations than running them over the edge of cliffs?) 8) Bison extirpation 1881-1882 grass height, density and fuel continuity ▲ Fire frequency ▲ (1881-1890 only) 9) Replacement of Bison with cattle and sheep ▲ grass height and fuel continuity ▼ Fire frequency ▼ woody fuels ▲

Table 4 attempts to summarize some of the relations between native wildlife and fire, as well as how changes in fire frequency may be related to European alterations of their numbers. Interpreting the table: The black triangles are arrows so to read the first line, Wolves ▲ means an increase in wolves and Coyotes ▼ means that as a consequence of the increase in wolves, the number of coyotes declines, and Prairie dog towns ▲ means that when coyotes (and other predators) drop, the number of prairie dogs can increase and Fire Frequency ▼ means that when prairie dogs increase there may be more prairie dog towns, which often lack sufficient fuel to carry fire and so portions of the landscape with many prairie dog towns may act as fire filters reducing the frequency of fire in the landscape.

HISTORICAL EFFECTS OF BEAVER ON FIRE FREQUENCY, SPECIES DIVERSITY AND POSSIBLE HABITAT FOR MOOSE AT CMR

Of the 21 major vegetation types at CMR (GIS map 1), most are close to their historical precursors but for some effects of fire suppression, livestock grazing and introduction of invasive exotic weeds such as Japanese brome (Bromus japonicus) and sweet clover (Melilotus officinalis). One exception is the

33 vegetation associated with the historical beaver wetland mosaic (Map 1, vegetation type 6.3) which were radically transformed by trapping, beginning in Montana by 1835, leading to collapse of dams in the mid 1800s, drying out, entrenchment of stream channels and diversion of water for irrigation beginning in the late 1880s.

Only a few streams at CMR had watersheds large enough that we can be certain about beaver, but long- collapsed beaver wetlands likely occupied the lower reaches of Armells Creek, Crooked Creek, Musselshell River and likely a number of the smaller streams historically. One significance of more beaver wetlands in the original landscape is that fire frequency would have been lower in the fire compartments between them. There may have been a little more habitat for fire refugial species such as Douglas fir because beaver firebreaks in the bottomlands would make it harder for fire to reach their cool ravines from below. The impoundments might have been expected to stop some fires ignited in the bottoms from spreading up onto the prairie uplands.

In both the southwestern and southeastern US, beaver may actually increase fire frequency by providing bridges for fire across small streams. There, beaver kill trees in moist bottomland hardwoods and swamps that would otherwise serve as good firebreaks. Then, after the trees die, the shallow ends of beaver ponds are colonized by species such as Scirpus cyperinus, Scirpus americanus (Schoenoplectus pungens) and other graminoids that produce a flammable thatch which persists beneath the green foliage, permitting fire to easily carry across small stream bottomlands. In Montana, the opposite appears to be true: in small stream bottomlands without beaver, such as lower Armells Creek today, dry prairie grasses and shrubs come right down to the channel and fire can cross easily. However, when beaver fill the bottomland with the usual string of small pools and connecting channels they create an ecosystem of poorly flammable cattails, Salix exigua thickets, small cottonwoods and pools that fire cannot cross readily unless there is enough wind to promote spotting. The result is that linear beaver ecosystems create fire compartment boundaries where none would exist without them, and, even though some percentage of fires may cross, downwind fire compartments will have, on average, a lower fire frequency. I would judge, however, that this would have had only a minor effect on overall fire frequency at CMR since most ignitions occur on the uplands.

Possible relation between extirpation of beaver and the decline of sharptail grouse? James Schultz claimed that sharptails that he saw in CMR were as abundant as they ever were in 1901. “For an hour or more after starting, we saw many flocks of chickens—sharptails of course—coming to the river for their morning drink” (Schultz 1902 p. 18). “As we rowed down past the wide Musselshell flat and through the rapids, we saw several flocks of chickens along the shore. Not coveys, but flocks of from twenty-five or thirty up to twice and thrice that number. They band together in large numbers at this season of the year [early November], and it was no uncommon sight to see several hundred of them at morning and evening winging their way across the river” (p. 81).

Before their extirpation, beaver wetlands may have supported larger populations of birds such as sharptail grouse that came down to them for water. During summer I noticed that the only place sharptails were conspicuous was in uplands with small streams that retained some water during July and August. Few such places remain today. I watched hens with broods of half-grown young foraging along creek margins under tall streamside cover. Beaver ponds in upper Armells Creek are bordered with tall cover of grasses, cattails and sandbar willow (Salix exigua) (Figure 23).

Bent 1932 (1963) observed that the range of sharptail grouse “is becoming more and more restricted as the Central West becomes more thickly settled and more land comes under cultivation.” Cultivation, however, could not explain the large decrease at CMR. Long before the farmer or rancher came the beaver trapper. Some of the preferred foods of sharptail grouse listed include buds of willow, leaves of cottonwood and in winter, buds of snowberry, all of which are common in and on the margins of beaver wetlands. Insects composed up to 95% of the diet of very young sharptails and these should be most abundant on wetland

34 margins. Nesting habitats include clumps of tall grass or shrubs on uplands and willow thickets and other cover on wetland margins. Within the refuge there may have been 100-300 miles of beaver-maintained wetland habitat that doesn’t exist now. It is interesting to speculate that the original small stream beaver wetlands may have been the vanished habitat for reproduction and rearing of young earlier in the year.

Ducks use the small beaver ponds along upper Armells Creek all summer and extensive beaver wetlands would have supported many more of them. They also may have maintained a higher diversity of aquatic and wetland plants. Good examples of a natural beaver wetland mosaic can be seen along highway 191 between Roy and Lewistown (Figure 23 below).

Figure 23. A natural beaver ecosystem retaining water in summer on upper Armells Creek between highway 191 and the Judith Mountains (background) the source of the Creek. Dominant vegetation is cattail and sandbar willow (Salix exigua) with small cottonwood sprouts, all of which provide an annually renewable crop of food for beaver, and habitat for a considerable diversity of submersed and emersed aquatic and wetlands forbs. There were ducks and plenty of water at time of photo in the dry weather of late July, 2007.

History of beaver trapping in the CMR area. There is much in the historical literature to suggest that beaver were a dominant feature in parts of the original bottomland landscapes of CMR before trapping reduced them to numbers too low to support their wetland mosaic. Trapping in the CMR dates to some time before 1840 when “free trappers’ (those not associated with any particular fur company), risked their scalps to roam the area. There were large beaver trapping expeditions to the south and west of CMR as early as 1833 (Larpenteur 1898). Trapping reached the Milk River by 1835 and the first beaver trapping in CMR probably occurred in that year. Trapping was patchy and isolated trappers short-lived until traders established intermittently peaceful relations with the Blackfeet around 1850. With construction of Fort Benton upstream in 1856 by the American Fur Company, trade multiplied. Schultz (1901) reported that hundreds of thousands of “wolf and beaver skins and pelts of the deer and elk were brought to it by Indian and white from the far North, from the South, from the Rockies and the vast extent of plains surrounding it, and were later shipped down the river to St. Louis.”

The ability of trappers to clean out an area in a few years should not be underestimated. In 1833 Charles Larpenteur, a greenhorn from the East, hired on to a beaver trapping expedition to virgin country in western Wyoming and southern Idaho. Trapping in the area called Pierre’s Hole near the head of the Green River, the party collected 10 packs of beaver in only 7 or 8 days. A “pack” of beaver skins weighed 100 pounds and consisted of about 60 beaver. On breaking camp to head for the Yellowstone River they had 30 packs or about 3,000 pounds equaling 1800 beaver on July 24, 1833. July 9 was the first day of beaver trapping so forty men, fanning out over the landscape trapped 1800 beaver in only 15 days! (Larpenteur 1898, p. 27- 31).

Although the Milk River and its tributaries “abounded” with beaver when visited by trappers in 1835 (Larpenteur) , Jim Bridger with a band of 30 men was disappointed in the numbers found when they trapped it in 1841. The system, however, had been trapped 6 years earlier, the word had been out and free trappers ranging upriver from Fort Union likely worked the area repeatedly in the five year interval. In 1857, however, a trapper named Dauphin had a successful winter in the vicinity, so beaver may have recovered in the 16 year interval since Bridger (Schultz p. 36)

By 1901 beaver were so scarce that trapping was illegal, but this did little to stop their continued exploitation. Schultz looked for beaver all along the river in 1901, not finding any until he reached CMR where he reported illegal trapping carried out by “French-Cree half-breeds”. “At the lower points of the island [Grand Island] I found some recent beaver cuttings, and also some moccasin tracks in the mud. From the shape of the latter I knew they were that of Cree make, and concluded that thee was a camp of Cree breeds somewhere in the vicinity. Alas for the beaver. They have been protected by law for a long time, but every year their number grows less and less” (Schultz p. 55).

Again at Two Calf Islands “There were several beaver slides on the lower one, but no fresh sign; evidently the moccasined trapper had been here also (p. 58). Dick King, a recent settler living on the next bottom below the mouth of Armells Creek (King Island) told him “Yes, they’ve pretty well cleaned up the beaver about here. I saw the old man floating by on a raft yesterday, and he had something covered with his blankets; beaver skins I suppose.”

Armell’s Creek heads in the Judith Mountains and has a year-round water supply, maintained in part by retention of water in beaver wetlands upstream. The same was true for the Musselshell River which has an enormous watershed, heading in the Lewis and Clark National Forest where its south fork drains the Crazy Mountains and the north fork originates in the Little Belt Mountains at an elevation over 7000 feet. The Musselshell also drains the Judith Mountains via Box Elder Creek, which flows into Flat Willow Creek and then into the Musselshell east of Winnett.

The American Fur Company ran a small trading post for beaver skins and other furs at the mouth of Flat Willow Creek around the late 1870s. “One time at our branch post on Flat Willow Creek we had been out of whiskey for some weeks, and Faval, who was camping and trapping nearby, was in despair because his large and ever-increasing pile of beaver skins could not purchase even a dram” (Schultz 1902, p. 70).

Downstream from this post, Schultz characterized Crooked Creek as a “dry creek” but it had water standing in pools in 1901, possibly abandoned former beaver habitat after some 30 years of trapping. By 1901 beaver had been nearly extirpated and it was illegal to trap them. Camping on an island at the mouth of the Musselshell in 1901 he found that “…several families of beaver have large caches of winter food at its lower end. May they escape the wiles of the trapper and increase. I would that it were life imprisonment to kill one of them.”

36 Beauchamps Creek and some of the other longer or wetter creeks and coulees originating from the Little Rockies may have provided water for beaver habitat, at least in their lower reaches. Shultz (1901) commented on several of these as “dry” creeks but mentioned that Beauchamp Creek was a “wet” creek with running water in wet seasons and with pools of water in the rest of the year and he had known it as far back as 1879. The pools again may have been remnant of beaver wetlands. The string-of-pools beaver ecosystems may have been breaking up then or had already collapsed earlier as the last of the beaver were being trapped. He also described two streams on the south side upriver beyond CMR as being dried up by 1901 by diversion of their water for agriculture.

Schultz (p. 92) again described Fourchette as a “running stream” lined with cottonwoods and willows along its course, and Kill Woman Creek a little further down as a dry stream. Seven Blackfoot also appears to have been dry, at least by 1901. Any of these stream bottoms that originally supported beaver would have been wetter than they are now prior to being trapped out after 1840. Armell’s establishment at Armell’s Creek was set up primarily for trading and trapping beaver skins which were still the most profitable fur until development of the felt hat in the mid 1800’s. Beaver had been nearly extirpated by 1901 and all of the creek mouths that I searched in July and August of 2006 and 2007 I would have called “dry” except for Armells, Musselshell River and Big Dry Creek.

Big Dry Creek has a much shorter watershed than the Musselshell but tends to retain some water all summer because for much of it’s length it lies near the Missouri River hydrologic baseline. The same is true of Willow Creek and one of its small tributaries, Beaver Creek, which originates in the Larb Hills and empties into Willow Creek only a mile or two from the refuge boundary near Fourth Ridge in Valley County. Both lower Beaver Creek and Willow Creek have highly meandering channels, productive grass bottomlands (1500 lbs per acre) and a water table high enough to sustain cottonwoods for some distance above the mouth of Willow Creek on the Milk River. As mentioned above, Milk River (presumably in references to its tributaries since it is too big to dam along its lower length), was said to “abound with beaver” when it was reached by trappers in 1835 (Larpenteur 1898). Another Beaver Creek originates in the Little Rocky Mountains, draining east and the north into Milk River at a point east of Bowdoin NWR.

Single trappers sometimes brought several hundred beaver skins out of a single watershed over a winter’s trapping. Such numbers imply the existence of nearly continuous beaver wetlands along the stream courses. After the large original populations were trapped out beaver were still a staple item at the trading posts as long as there were still a few to be had. Sometime between 1877 and 1882 Schultz worked at a fur trading post on CMR at Carroll bottom where one winter he mentions that they took in 300 beaver skins. This was a far cry from the quantities traded from 1833 through the 1860s, after which the buffalo robe became the predominant article of trade.

Beaver activities—building dams and retaining water—facilitating the slow release of water through the dry season, have a positive feedback on the numbers of beaver until their potential habitat is all built. After that, if the water supply meets minimal requirements, they are capable of maintaining their irrigated wetlands on a permanent basis. Beaver impounded water streams like Armells Creek that now runs out of the system in days. The shallows around the margins and in the headwaters of beaver ponds are colonized by cattails and willows which, along with small cottonwood sprouts and a few shrubs, constitute their future food supply. As such, beaver are agricultural animals, creating an irrigated system and then cropping the annual food produced. The more beaver wetland created, the longer the water was retained after spring snowmelt and summer rain events. As a result these systems acted as sponges, slowly releasing water from one pond to the next below, and certain of the local streams should have been sustained as permanent wetlands from the point in the headwaters where tributaries first assembled sufficient water, all the way to their mouths. These systems, lying in the lowest and coolest parts of the landscape, would not have been expected to dry up. Even tiny stock ponds, dug into the highest and driest parts of the landscape rarely dry up in summer.

37 Beaver are ancient animals, dating back to the Eocene, some 49 million years ago. In the CMR area they would have been present for 6000 or more years, dating to shortly after the retreat of the last vestige of the Wisconsin glaciation ice front at CMR. That would be enough time for geologic accretion of sediments in beaver bottoms to create the flat bottom profiles seen in many of the small tributary streams. These are the flats into which the small streams have now become entrenched (compare Figure 23 above on upper Armells Creek and Figure 24 below. It seems likely that they maintained a now-collapsed wetland system along at least three major streams and the lower ends of several minor streams at CMR.

There are historical records of beaver system collapse after trapping. In Arizona in the late 1800’s the US Army carried out a program of trapping beaver in order to drain all the cienegas—desert wetlands created by beaver along small streams—to eliminate the mosquito that carried yellow fever. On the San Pedro river south of Tucson long-time ranchers described the rapid disappearance of once extensive wetlands after extermination of beaver. Without the dam complex, the system rapidly collapsed and the San Pedro, a two-foot-wide trickle that I stepped across at this location in 2003, reportedly cut a deep channel into the soft sediments of millennia during runoff events. One account in 1901 by C. H. Bayless, owner of a large ranch near Oracle (north of Tucson) since 1886 said that after trapping the beaver in the 1890s, the bottomlands drained and “…within four or five years a channel varying in depth from 3 to 20 feet deep was cut almost the whole length of the river” (Bahre 1991). Today only about 12 small remnants of cienaga remain in Arizona; many contain rare wetland plants and the remaining habitats are the subject of protection efforts.

It is tempting to speculate that something similar happened at CMR. Strom (1984) commented that “the potential native plant community (on some Fluvaquents in McCone County) has been altered by the entrenchment of a creek in the soils”. Deep cuts through formerly deep alluvium can be seen in a number of small stream bottomlands such as Armell’s Creek. The beaver site in Figure 24 below is near the location of the American Fur Company’s trading post operated by Charles Armell around the 1860s, some years before Schultz’s arrival in1877 (Schultz 1901, p. 58). The early efforts of the company were focused on “fine furs” of which beaver headed the list, and it seems plausible that beaver were trapped out of lower Armell’s Creek before the Civil War. In addition to trapping, much water from the upper watershed of Armell’s Creek was used for agriculture by 1900. In that year James Fergus advertized that his 10,000 acres included 30 miles of irrigation ditches in the upper Armells Creek watershed, diverting water for oats, hay fields and stock watering.

Figure 24. An altered beaver system with a small beaver impoundment in the bottom of the entrenched Armell’s Creek just west of US 191 and south of the bridge. Given the antiquity of beaver in the area, their near extirpation by trapping and the history elsewhere of stream entrenchment after removal of beaver, it may well be that all of the bottomland in this photo was once beaver wetland. If true, we can speculate that the existing channel is an artifact of post-beaver entrenchment, about six feet deep here; the green area would correspond to the deeper water of beaver pools or flooded channel, and the silver sage has colonized fine sediment in the shallows of former ponds once filled with cattails and sandbar willow. The small pool, barely visible in the channel, is supported by a recent small dam. Beaver here may be starting to rebuild the system by trapping sediment to fill the erosion gullies, a process that could take many decades or hundreds of years even if they are not trapped. Documenting this process might be a useful activity of the refuge.

Diversion of small stream water supplies for irrigation. By 1900 there was government enthusiasm for the greening of the prairie through irrigation (Bureau of Agriculture 1900) and some areas around the Judiths and further upstream had already been irrigated for around 10 years. Below Benton, passing the mouth of Shonkin Creek, which originates in the Highwood Mountains, Schultz noted in 1901, “It is a stream no longer. Once it was a good-sized creek of pure mountain water. Schools of trout lived in its clear depths, and the beavers bridged it with their dams. Then came the white man and used the water to irrigate vast tracts of the barren plain, so nothing now runs in the old channel but a little seepage of brown alkaline water. The trout are dead, the beavers have vanished, never to return.” Again, tramping along the edge of Arrow Creek, above the mouth of the Judith River: “Water was standing in pools here and there in the creek bed. The ranchers away up in the Judith basin have long since diverted Arrow Creek to irrigate their homesteads and it is no longer a running stream except in the June rains.”

With beaver largely trapped out at an early date and ranchers like James Fergus diverting the flow into his 30 miles of irrigation ditches, the elimination of former beaver wetlands in Armell’s Creek in not an unreasonable picture. If so, all this occurred even before the big influx of homesteaders, leaving lower Armells dry now for over a hundred years.

Possible moose habitat maintained by beaver. On May 10, 1805, John Ordway of the Lewis and Clark expedition said that the hunters saw moose around 10 miles by the loops of the river upstream from “Big Dry River”. They had passed that point and camped at the mouth of present day Duck Creek on the north

39 side the night before. They made only about four miles on the tenth before stopping to camp and send out hunters. “The hunters killed a fat buffaloe 4 Beaver & 2 black tailed deer and one white taild deer. they Saw Several moose deer which was much larger than the common deer and the first we have Seen” (John Ordway’s journal 1805 p. 146).

Moulton, editing Ordways’ journals, adds a footnote suggesting that Ordway must have been mistaking mule deer for moose, but the fact that Ordway distinguishes “black-tailed” deer, whitetails and a group of much larger “moose deer” in the same sentence suggests that the hunters knew what they were seeing. This is supported by the historical place name “Moose Point” on the north side a little downstream and by other historical sightings. Records of early beaver trapping suggest the possibility that beaver-maintained moose habitat may have been eliminated by drainage during the period 1840-1860.

It may be significant that Orway reported “several” moose, as opposed to a single that might have been a wandering stray from Canada. The camp may have been in the vicinity of Skunk Coulee but Ordway does not say which way the hunters went. If they went up Skunk Coulee they may have seen moose in the Willow Creek bottomlands. Ten miles downstream from Big Dry Creek, and about 15 miles from the moose sighting of 1805 was the neck of land near the original site of Fort Peck called Moose Point (Larpenteur 1898). Moose Point, named in the earliest times, was only one ridge south of Willow Creek and the virgin bottomlands of Big Dry Creek and Willow Creek might have supported moose year-round if extensive wetlands were sustained by beaver activity. If there had been some amount of permanent habitat for moose maintained by beaver and resident moose in this vicinity, beaver trapping should have eliminated the habitat by the 1840s even if the moose themselves were not killed by hunters and trappers.

Future work. The comments above are only notes from skimming the surface of the historical literature for clues to the importance of beaver in the original CMR landscape. Possible wildlife management implications of beaver wetlands, such as an original source of drinking water for bison when they were far from the big rivers, or the possibility of critical nesting and brood-rearing habitat for sharptail grouse, hinted at above, suggest the need for more work to define the original extent of beaver wetlands. Far more information will likely be found in the historical records of the American Fur Company and others involved in the 50 years of intensive beaver trapping in the CMR. Also, I saw strata of fine woody material in the cut sides of some of the small stream bottomlands in Phillips County where stream channel entrenchment has taken place. It would be worth a more detailed study, including assembling more historical information, and investigating sediments along the now entrenched channels of the major tributaries to look for deposits of fine sediment and plant fragments characteristic of former beaver wetlands or any preserved wood from ancient dams.

EFFECTS OF PRAIRIE DOGS ON FIRE FREQUENCY

An attempt was made to assess the historical literature and to study the soils preferred or rejected by prairie dogs at CMR in order to interpret their effect, if any, on historical fire frequency. Soils, moisture and slope characteristics were used to rank all soils at CMR in terms of preference or rejection by prairie dogs and we made a GIS layer to show their likely pre-european settlement habitat.

In the pre-european landscape the shortgrass prairie was dotted with prairie dog towns and prairie dogs had existed for thousands of years in some sort of equilibrium with their natural diseases and predators. Viewed from the air the colonies usually have rounded shapes with smooth boundaries. Fire rarely burns them because the prairie dogs eat the potential grass fuel within the town and cut down shrubs to eliminate hiding places for predators. Like chicken, everything eats prairie dogs. Living and raising their young in the same burrows, the endangered black-footed ferret was the most immediate predator. Outside on the surface prairie dogs have to be on constant lookout for danger from the air in the form of eagles, harriers

40 and other hawks. Badgers are a threat where coarser soils are easy to dig. The town margins are tension zones: the prairie dogs are attracted there where grasses are more lush but behind any large sagebrush may lurk a bobcat or coyote. Abandoned burrows are common around the margins.

Along with barren soils, badlands and rock outcrops, prairie dog towns constitute one of several elements of patchiness in the fuel landscape. A single town is an island which does not stop fires, which simply flow around it, but a landscape locally saturated with prairie dog towns could serve as a firebreak or fire filter, reducing the number of fires that make it through and reducing fire frequency on the downwind side.

In 1805, the Lewis and Clark expedition encountered a prairie dog colony at the mouth of Sand Creek: “2 ½ mile up [from the mouth of Rock Creek] a creek falls in on the Lard. Side opposite a large village of Barking Squerrells” (Ordway 1805, p. 154). It is not clear whether he meant on the opposite side of the river or just next to the mouth of Sand Creek but on its upriver side there is a large body of Marvan silty clay, a preferred soil for prairie dogs, so this historic location could be added to the list of known sites at CMR.

A few miles down river from Seven Blackfoot Creek, James Schultz encountered the only prairie dog town he mentioned in the bottoms of CMR, on a flat, now inundated, below a formation called the Sphinx. “We crossed the flat, passing through a prairie dog town, where the little animals were so tame that that they sat up on their mounds within fifteen or twenty yards of us, an scolded us unmercifully. Evidently, they knew nothing about men and rifles. We left them, still barking and jerking their tails, and began the ascent of the valley slope west of the little creek” (Schultz 1901 p. 109)

41 Figure 25. Prairie dog town (yellowish patch) on southernmost upland flat of U.L. Bend, Phillips County, unburned in an otherwise completely burned landscape. Portions of the Soda Creek Fire can be seen across the river in the background.

There is no hard evidence that prairie dogs were ever very significant in terms of fire filters in Fergus or Petroleum County portions of CMR. There is too much topography. On uplands only a few polygons of FE233 (one PDT at Sand Creek on 233) and a few small pockets on FE 234 (one near Sand Creek office and one near intersection of Wilder Trail and Sand Creek trail) had documented examples of prairie dog towns and seem open enough, flat enough and large enough to appeal to prairie dogs. The rest of known locations were on the drier clay terraces along the Missouri River where they would have had little effect on fire flow. Prairie dog habitat in these bottomland sites seems limited to soils too high above the water table to support cottonwood forest or dense tall wet bottomland grasses and flat enough to provide their preferred gentle, convex topography, and open enough to prevent hiding places for predators. See in particular soils such as FE140, 141Kobar silty clay loam and FE166 Marvan silty clay.

In contrast, portions of Phillips, Garfield and McCone counties that have soils and topography preferred by prairie dogs are occupied by remnant colonies.

Prairie dogs may have inhabited the original landscape in large numbers and several authors have attempted to estimate the presettlement numbers of prairie dogs. The question for our purposes is, “Was the density of prairie dog colonies in any part of the landscape high enough that the nearly fuel-free clearings they created reduced natural fire frequency?” In the Montana Archives at Helena I looked at some 70 historical photos that showed some portion of natural prairie. The photos, mostly from the Haynes collection, were taken in and around CMR and other places in eastern Montana around 1880, while bison were still abundant locally, definitely pre-European settlement and before the introduction of plague from Europe. There was no sign of prairie dog mounds in any of the photos. It is certain that they did not blanket the landscape but had preferred habitats.

Mapping Historical Extent of Prairie Dog Towns as a Step in Fire Frequency Mapping All 278 soil units at CMR were evaluated for suitability for prairie dog use. Two types of study plots were done at sites of prairie dog towns: intensive 1/10 hectare vegetation plots were done at a number of definitive sites, while quick 1 page evaluations that included soil type, slope, landscape position and dominant vegetation were recorded at others.

Prairie dog habitats are identifiable by some nine soil and landscape features: • soil texture: the densest clay soils are preferred – see table below. Sands and fine sandy loams are avoided, probably because of the ease with which a badger could dig into burrows in the looser soils. • soil moisture: soils seasonally flooded or pooled are avoided. Nevertheless, Schultz reported one instance where a prairie dog town was drowned by deep water associated with a spring ice dam in the Missouri River bottomland. • slope: the flatter the better, or rather the more gently concave the better. Imagine if you wanted to be able to see everybody in your town from the dirt mound on your doorstep. With one or two exceptions all occupied sites at CMR were on gently concave sites. When a colony is expanding convex sites, such as those on rounded ridgetops are avoided, perhaps because barking for predator warnings could only come from the occupied side, leaving the other exposed. Line of sight also becomes a problem on a convex site: if the prairie dog is on one side it cannot see a predator coming up the other side. In the exceptions, colonies were seen on convex hilltops only where both of the concave lower sides had already been colonized, providing warnings from both directions. • The steepest habitat for prairie dogs was seen on a slope of about 15 degrees on toe slopes of a small hill on the east side of the Haxby Point peninsula. All other sites had lower slopes.

42 • topography: empty prairie dog holes, identifiable by cobwebs across the entrance, were common on the margins of PDTs where they approached within 30 feet of abrupt dropoffs and gullies that could hide predators. • Landscape position – some soils with preferred texture were avoided because they occurred only on unsuitable narrow ridgetops • Soil productivity (grasses) – the most productive soils (>1500 pounds per acre) were avoided. Tall, dense grasses would provide stalking habitat for predators. • Trees and shrubs: sits with any amount of woody cover were avoided. • Rock outcrops – sites near outcrops which could conceal predators were avoided.

Site characteristics preferred by prairie dogs at CMR included the heavier soils: 1) soils in order of frequency used (for soils used more than one time): 11 clay and gravely clay 13 clay loam 9 silty clay 8 loam 6 silty clay loam 6 silt loam 6 fine sandy loam

Preferred Soils, 21 series used more than once: 7 Gerdrum clay loam 4 Absher clay loam 4 Bascovy silty clay 4 Twilight fine sandy loam 4 Vanda clay 3 Cabbart silt loam 3 Cambeth silt loam 3 Elloam gravely clay 3 Kobar silty clay 2 Chinook fine sandy loam 2 Eapa loam 2 Evanston loam 2 Kobase silty clay loam 2 Marvan silty clay 2 Meganot silty clay loam 2 Nobe clay 2 Sonnett silty clay loam 2 Sunburst clay loam 2 Thebo clay 2 Yamac loam 2 Yamacall loam

Table 5. In all there were 73 soil series included within the 43 soil map groupings that were used by prairie dogs at CMR. This table shows the preferred soils.

Where prairie dogs were present, soils were classified into four categories of utilization and extrapolated to the rest of the 278 soils so that all soils at CMR were ranked and this was used for the GIS map:

1 Highly preferred soil type, as indicated by presence of large towns in multiple locations. Mostly clay or clay loam soils

43 2 Moderately occupied soil type preferred (utilized), as indicated by presence of entire or parts of prairie dog towns, often spilling over from the main colony on an adjoining, preferred soil type. 3 Marginal soil type, mostly occurring as spillover areas onto adjacent non-preferred soils when all of the preferred soil type has been occupied. Usually of small extent. Occasionally on the toes of slopes of soil type mapped as slope class E or even F which are never preferred soils. r = Rejected soil types, often because of wetness, slope, or woody vegetation but also sometimes on soils that were flat and extensive but clearly rejected when the colony limits stop abruptly at the new soil type. In one case a large town on a Gerdrum clay loam stopped along the edge of a Busby-Twilight fine sandy loam (McCone county, soil map 78 at Nelson Creek). In this particular case the sandy loam was occupied by a tall stand of needle-and-thread, dense enough to provide cover for predators and with light, loose soil, soft enough to make for easy digging for badgers, one of which was seen nearby. Sandy loams and sands were rarely utilized even when topography seemed ideal (6 fine sandy loams and 1 loamy fine sand were used).

There were 278 soil types used by NRCS for mapping at CMR This includes types mapped separately for different slopes classes of the same series and sometimes the same pair or trio of series in different combinations, such as Phillips-Elloam and Eloam-Phillips-Absher. These combinations were developed from 87 discrete soil series. Sixty six of the 278 mapped units were found to be clearly occupied or clearly rejected by prairie dogs. Of these, 43 types were used by prairie dogs and 23 were rejected.

The remaining 207 soil map units not found in the immediate proximity of prairie dog towns were evaluated and scored for suitability for prairie dogs. Each soil was evaluated on the basis of wetness, slope, soil texture, amount of tree and shrub cover, rangeland productivity figures, position in the landscape and continuity of grass as interpreted on aerial photographs. Many units could be easily excluded by being too wet, too rugged or too wooded.

The reported historical prairie dog increase noted by people in various parts of their range to have followed the disappearance of the buffalo was likely the result, not of any change in buffalo numbers, but the elimination of all but one of their predators, the black-footed ferret. A possible prairie dog peak, still remembered by some older people, ended around the 1930’s in the region with intensive public and private poisoning campaigns, followed by introduction of plague.

George W. Wingate on a trip through the Yellowstone in 1885.(Wingate 1886) noted “The wolf and the coyote were once extremely numerous in all this section of the country, but they have been nearly exterminated by poisoned carcasses prepared for them by the cattlemen to whom they cause a great loss by killing the young calves. It is said that in consequence of their destruction the prairie dogs are increasing greatly” “We saw but one coyote, and heard none, nor any wolves during our entire journey.” p 228 “The wolf hunters and cattlemen wage unrelenting war upon all wolves and coyotes, and strychnined carcasses are so strewn over the prairies that it is never certain when a dog will be poisoned.”

The poisoning of wolves, coyotes and collateral damage to other prairie dog predators who might feed on the poisoned carcasses peaked around the same time that the buffalo declined and prairie dogs increased, almost surely because of the elimination of some key predators. .

There are a number of other historical reports of prairie dog increase in other areas after disappearance of bison., so that, while not documented quantitatively, it is hard to dismiss what people say they saw. In most areas the extirpation of bison coincided with increasing poisoning of wolves, coyotes and other predators so there is little reason to suspect that bison actually had anything to do with any increase. The likelihood of local reduction of fire frequency by any naturally dense prairie dog towns was one of the factors used in assigning first approximation fire frequency to each of the 134 fire compartments mapped at CMR (below).

Effects of Bison on Fire Frequency

Bison numbers. How do early 20th century and modern numbers of grazing animals compare with the historical numbers of bison? We still have only rough estimates but McHugh (1972) estimated a maximum number of buffalo of 30 million, using figures for grassland productivity and carrying capacity and considered the great difference in carrying capacity between tallgrass, shortgrass and other habitats. His estimate seems to have covered the entire historical range of bison from Florida to Alaska. The figure included both grasslands and wooded areas and he subtracted acres to compensate for competition from other grazers such as elk, deer and pronghorn. In his map there are 17 states and provinces in the grassland “core range”. The core states and provinces include large parts of Saskatchewan and most of Montana with exception of some of the mountains of the western part. Rangewide, subtracting the slightly less than 2 million animals in wooded portions of the range leaves around 28 million in grasslands. 28 million divided among the 17 core states equals about 1.6 million per state. It should be possible to do better than this but I have not seen a study specifically focused on numbers of bison in Montana.

If McHugh’s rough calculation happens to be accurate for Montana then subtracting 1.6 million bison from 2.7 million cattle & sheep today less equals an extra 1.1 million more animals, or about 70% more than in pre-European times (67% if we ignore sheep and horses). That is today: in the decade 1900-1910 there were an average of 5.2 million cattle and 1.1 million sheep. Comparing only cattle that would be 5.5-1.6 = 3.9 million extra animals or more than triple the numbers of bison in the pre-1882 landscape. Add to that 1.1 million sheep and many thousands of horses would make it quadruple the original number of bison.

None of this allows for the difference in size of animals or the different rate of utilization between cattle and bison. Even disregarding sheep and horses and allowing for 1 bison to be the equivalent of 1.2 or 1.3 cattle still indicates triple the grazing pressure on the land in the early 20th century. This picture suggests near elimination grass fuel to carry fire and grazers and agrees with the doubling of the length of fire return intervals in the first half of the 20th century seen in the fire scar chronologies. And it is not surprising that this was the time that the understories of wooded areas began to fill in with tree saplings and shrubs (see the tree demography studies below).

Figure 26. Bison on the Theodore National Wildlife Refuge In western North Dakota

Some startling results on the effect of bison grazing in shortgrass prairie at CMR appear in the Fire Scar Chronology section below.

HISTORY OF CATTLE AND SHEEP GRAZING IN THE CMR REGION

In 1833 Charles Larpenteur helped drive six cattle (four cows, two of which were milk cows, and 2 bulls) on a 3000 mile loop beginning along the south side of the Platte River, across the prairies of Nebraska and Wyoming and from Laramie River to Pierre’s Hole, which was just into southern Idaho and southwest of Yellowstone. While there, the group of 40 men trapped some 1800 beaver in just 15 days. On one of the last days a bull was bitten by a rabid wolf and died of hydrophobia. The remaining bull and four cows were driven back across the continental divide into Montana where they may have been the first of their kind to set hoof into the state. “Our two cows added a great deal to our good living; as we had no coffee, milk was a great relish” (Larpenteur 1898 p. 39). The party followed the Yellowstone River across the central and eastern part of the state to their destination at the mouth of the Yellowstone, the intended site of Fort William. The day after arrival, the men commenced building the fort just over the North Dakota border about 3 miles downstream from the competing trading post at Fort Union on the Montana line (Larpenteur p. 35, 39). After this introduction cattle seemed to have been kept here permanently. Fort Union maintained a hay field beginning in 1834 despite occasional Indian attacks, bringing in hay to see the stock through the winter each year. There seemed to have been a small herd maintained close by the fort for milk and ox power. In 1844 Larpenteur reported that Indians chased away some of the milk cows from Fort Union and shot them at some distance from the fort. Cattle were easily killed by Indians long accustomed to hunting buffalo so there was no possibility of free ranging cattle as long as there were Indians. Again in 1847 a war party of Assiniboine Sioux attacked the men at the hay field, wounding one and killing four oxen hitched to hay carts.

Sometime between 1832 and the early 1840’s cattle were taken upriver to Fort McKenzie (1832-1844) and then, after burning of the fort in 1844, down to Fort Chardon at the mouth of the Judith River. These became the first cattle in central Montana, but were still closely held about the forts. Fort McKenzie had at least one pig, as one was reported killed by the Blackfeet in 1844. Fort Benton (1847-1864 and the competing Fort Lewis (1845-1847) would also have had cattle. By 1860 cattle were a regular part of many trading outfits. Larpenteur crossed the plains of North Dakota to Fort Stewart on the Missouri near the mouth of Poplar River (in eastern Montana halfway between Fort Union and Milk River) overland with eight wagons drawn by oxen (p. 265). In 1861 he sold his cattle, around 10 or 20 head, and wagons at Fort Stewart to “some gentlemen who resided in Bitterroot Valley, one of them a merchant named Warren…” p. 276-278. This began the era of establishment of permanent local cattle herds in the headwaters of the Missouri River in areas where the Indians had been driven out. That fall he bought 20 more cattle at St. Paul and used them to pull more wagons overland across the prairies of North Dakota to Fort Stewart, the next fort downstream from Milk River at the mouth of Poplar River. This herd all perished during the severe winter of 1861-1862 at Fort Stewart. Shallow water on the Missouri often stopped steamboats from reaching the Fort Benton area so by 1863 freighting goods with ox and mule-drawn wagons from Milk River, Carroll, Rocky Point, Judith River and other landings along the river became common practice.

White European populations and ranches expanded after each of a series of treaties and non-negotiated Congressional boundary decrees pressed the Indians away from settled areas. By the late 1870s, Texas longhorn cattle were being driven up into central Montana and in 1880 much of central Montana from Helena to Benton and east to Lewistown and the lands around the Judiths was becoming reasonably safe for cattle ranching. The major Indian groups in the CMR area were the Blackfeet, composed of three divisions: the Blackfeet (north of the Little Rockies, to the Saskatchewan River in Canada, and originally as far north as Edmonton), the Bloods and the Piegans (sometimes called the Pikunis), the southernmost group who hunted the whole of CMR, often spending winters on the Marias River and summers in the vicinity of

46 the Musselshell. They interacted with the Flatheads further west, whose influence did not extend to CMR, and with the Assiniboines to the east and the Crees to the southeast who did range into the Musselshell River area, the young men of these three tribes often taking war parties to raid and steal horses from the others.

Cattle herds in the Helena valley multiplied rapidly from 1865 on. James Fergus began acquiring land and cattle there after leaving the 1862 Fisk Northern Overland Expedition which explored some of the route later taken by the Great Northern railroad while on their way to the gold fields further west. Cattle were being driven in numbers into the lands surrounding the Judiths even before the final collapse of the buffalo and removal of Indians. In the 1880’s a stockman’s landmark was the Musselshell Crossing east of Roundup, where herds of Texas longhorns, driven north, were bedded down for the last time before being fanned out in smaller herds to their ultimate Montana owners. In 1880 it had a store and in 1883 a post office (Cheney 1983).

The era of the big ranches, 1880-1910. In 1880 James Fergus, the “father of Fergus County” began acquiring land for cattle and later sheep ranching on the headwaters and tributaries of Armell’s Creek. After becoming prosperous at Helena from 1865 to 1880, he moved his herd of 900 cattle and 100 horses from Helena to Armells and he, his son Andrew and his brother William, over the two and a half decades 1880-1905, participated in conversion of the land from Indians and buffalo to cowboys and cattle.

For the first few years after 1880, cattle on open range grazed far from the land actually owned at Armells. In the eighties the range of the Fergus brand, the F Bar, was the vast expanse from the Judith River on the west to the Musselshell River on the east, and from the Judith Mountains on the south to the Missouri River on the north. This Moccasin and Cone Butte Association range covered about 2,400 square miles. Thus Fergus and son took over all of the former Blackfeet buffalo hunting range south of the Missouri scarcely before the buffalo were gone and the Blackfeet confined to lands on the Marias River.

As soon as the last buffalo were slaughtered vast numbers of livestock were pouring into the lands around the Judiths. Andrew Fergus’s large sheep ranch on Box Elder Creek was established shortly after James, in 1883. In that year also, 6,100 sheep were driven in from Oregon by Oscar Stephens and a total of 33,315 sheep were reported to be in the Armells area (Willmore, 1987).

Some time shortly before, Granville Stuart, a friend of Fergus, had brought in some 5000 head of Texas longhorns. The DHS ranch on the south side of the of the Judiths, the Maginnis Range, was of similar size. The headquarters of DHS (Davis, Hauser and Stuart) was at the Burnett Ranch (Willmore, 1987). The equally large C.K. Company apparently held range lands on the north side of the Missouri and east toward Glendive. Thus the range occupied by the buffalo was apportioned among a few large ranching outfits immediately after their disappearance.

By 1892 James Fergus and son expanded their herd of 900 range cattle to 4000 and in 1894 they got into sheep in addition to cattle. He and his son Andrew formed the Fergus Livestock and Land Company in 1895, and beginning with the Armells homestead and desert land claims in the early 1880's, the ranch eventually grew to 8600 acres of deeded land with water rights, plus 2500 acres rented from the state. James noted in 1897, "We have about 4000 acres, all on the streams. By controlling and fencing up the water, we control the range. Water is worth more here than land." [James Fergus to Judge Richardson, September 3, 1897, Box 21 F. 4, Scrapbook, p. 97, FP, UM. They were able to simply control open range by dominating its lifeline of water.

The largest Fergus purchase was an entire ranch in 1896, Eugene Townsend's 1000 acres with all equipment and 4000 sheep, 60 horses, 125 cattle. In a separate operation, William Fergus amassed from 1883 to his death in 1905, 8000 acres and 20,000 sheep.

47 In 1896 Fergus expressed regret with the company's decision to increase their cattle holdings because of a "crowded range”. This is in contrast with the glowing accounts in the state’s 1900 propaganda campaign to bring in homesteaders. In 1901 James noted that “all our income comes from our sheep.” In 1893 he had attempted to sell 600 range horses, which he had come to think of as a liability but the nation was going through a major depression at that time and they failed to sell. By 1900 James was 86 and had been thinking about selling out for some time. The company had by that time 2500 cattle valued at $75,000, 1000 horses worth $15,000, and 9000 sheep at about $30,000. He posted the following ad that year.

FOR SALE On account of old age and ill-health, will sell about 10,000 acres of land in Fergus county, Montana, nearly all on tributaries of Armells creek, on the north side of the Judith mountains; has 30 miles of irrigation ditches, 80 miles of fences, nine homesteads or sets of ranch buildings, together with about 2,000 good cattle, 9,000 well bred sheep and 1,200 horses, originally bred from grandsons of Mambrino Patchen and other good stallions; 6,000 bushels of oats, several hundred tons of hay, two blacksmith shops; one-third interest in 55 miles of the phone line and some private; 200 acres in crop this spring; an interest in sheep shearing sheds and apparatus, and all the necessary implements for running such a ranch, including a post office. Fergus Livestock and Land Co. James Fergus, President, Armells, Fergus County, Montana. [James Fergus, advertisement sent to unidentified paper, 1900, Box 21 F. 5, FP, UM. Fergus advertised in the Drover's Journal, Box 19 F. 44, and the Chicago Daily Journal, Box 2 F. 40, and probably in Canadian papers such as the Calgary Herald, since he had previously advertised horses there. Fergus Memo, n.d. (late 1890's), Box 14 F. 4, FP, UM.]

James Fergus died in 1902, age 87. In 1900 he had complained that the ranch was slowly failing financially and among the reasons he enumerated for selling was the overgrazed condition. Figures 27 and 28 give some impression of the range “fed off” by sheep grazing.

Figure 27. Loading wool for market from James Fergus’ ranch at Armells in the 1890s. In the original photo which has more detail, the foreground and hills in the background can be seen to have been grazed bare. It was around this time that Fergus described the range condition as having been “fed off”.

48 All of the Fergus information above is from James Fergus Ledgers and Daybook, 1880-1902 and other correspondence and papers in the Stuart Collection, Montana Historical Society Library, available online at http://www.dangel.net/AMERICA/Fergus/James/CHAPTER%20XIII.html.

By 1900 sheep had outpaced cattle as the leading industry in Fergus County, producing mutton and four million pounds of wool (Bureau of Agriculture 1900) and much of the prairie ranges was in dire condition. After saturation of the best lands by large cattle and sheep operations, the next wave of exploitation of the former buffalo range consisted of filling out the more marginal landscapes by smaller operators. In 1901 James Schultz and his Indian wife Sahnéto took a float trip down the Missouri from Benton, through CMR to the mouth of Milk River to revisit places where Sah-né-to had grown up following the buffalo with the Piegan band and where Schultz had worked as a trader in buffalo robes and other furs beginning in 1877.

Figure 28 Sheep on overgrazed winter range in the Yellowstone breaks, 1907. L.A. Huffman photo courtesy of the Montana Historical Society Library.

Schultz was disgusted with what sheep were doing to the land. At the end of their first day, 42 miles below Benton, he and Sah-né-to camped at the Coal Banks. “I remembered that on my last trip down the river in April 1882, we had camped for the night in a narrow strip of cottonwood and willows, and thither I directed Sah-né-to to steer the boat. We found on landing that we were within the bounds of an accursed sheep ranch; but in memory of old times I decided to camp there anyhow,…”

A second ranch was encountered at the mouth of the Judith River where William Norris had a ferry, general store and several hundred acres “irrigated by a ditch from the Judith” (p. 31). Ranch three was a mile below Norris on north side of the river “…which depended upon a wheel for irrigation. It was an immense affair of wood and steel rods, sixty feet in diameter, and revolving by the force of the current against its

49 broad blades. Large, deep troughs, or buckets, took up the water and poured it into a long flume extending to the irrigated land. It kept up a constant stream of more than 100 inches and that quantity will water a large acreage.”

A few miles below the mouth of the Judith River they encountered a group of men building a cabin at Dauphin Rapids. He asked them “Are you building a sheep ranch?” They replied “Not on your life. We’ve got a little bunch of cattle; the sheep men run us out where we were located over on the railroad, and we’ve found a good range here. The first blankety-blank sheep man that shows up in this vicinity with his flocks had better come heeled, for we’ll sure fight.” Schultz added “The sheep men are, without doubt, killing the golden goose; the luxuriant range which would have lasted forever if stocked with cattle only is being rapidly ruined by them” (Schultz p 35),

The fifth ranch mentioned by Schultz was a small cattle operation owned by his former neighbor Dick King on the next bottom below the mouth of Armell’s Creek, now called King Island. With cattle only established the previous year in 1900, King told them that they found no need to work and store up hay for the winter: “Here we don’t need any. Cattle find ample feed and shelter here in these bottoms and keep fat during the worst of winters.”

The sixth ranch was owned by his friend Mark Frost. This seems to have been on the north side of the river at the beginning of the southward bend from Hawley’s Flat to the mouth of the Musselshell. The cabin was situated on a cottonwood flat with abundant whitetail deer, now all flooded by Fort Peck Lake. The flat was about an hour’s run drifting in a wooden boat from the mouth of Beauchamps Creek. That would put it in the vicinity of the end of refuge road 219 which comes down to the river from Hawley’s Flat which would have been Frost’s grazing area. Having been located on the river “for a number of years” in 1901 that would put its establishment around perhaps 1895-1898. “He has a fine ranch, a nice band of cattle, which support themselves the year around, and so has absolutely nothing to do but enjoy himself. Once a year he gathers his beef stock and drives them to the railroad ninety miles distant [the Great Northern at Malta, a major shipping point for livestock, which in 1899 loaded 18,000 cattle, 20,000 sheep and a million pounds of wool (State Bureau of Agriculture 1900)], ships them to Chicago, and purchases a year’s supplies, and then he goes back to his ranch.”

The seventh and last ranch encountered was in the river bottomland below Herman’s Point on the south side of the river from the mouth of Kill Woman Creek. “Opposite the mouth of the little creek we saw a skiff tied to a stake, and going ashore beside it, I climbed the trail leading to the top of the high bank. Just as I reached the top I nearly ran into a man coming after a bucket of water, and was not a little surprised to recognize my old friend Ed Herman. ‘Well, well’, he exclaimed, almost yanking me off my feet. ‘What on earth are you doing here?’ ‘Oh’ I replied, ‘just drifting along and revisiting our old stamping grounds with Sah-né-to. And you?’ ‘Why, I’m building here and going to buy a few cattle. There’s my shack over there; just bring up your bedding and make yourself comfortable until next spring at least’ ”

Herman had seen mountain lion and many bighorn sheep in the rugged breaks near his new place. The house was on a river flat, now inundated like that of Mark Frost upstream and below Herman Ridge, on the south side of the river, presumably named for him.

With exception of the first ranch, a sheep operation and Norris’s store and ranch at the mouth of the Judith River, all of these small cattle ranches in the river bottoms and breaks seem to have been just established, in 1901 (the group just starting up downriver from the sheep ranch), 1900 (Dick King), ~1896 (Mark Frost), and 1901 (Ed Herman). All the cabins were on cottonwood flats in the river bottom, with the last two flooded out by Fort Peck Lake.

50 Traveling overland in 1880 on horseback from Kipp’s trading post near Carroll, over Crooked Creek to their branch post on Flat Willow Creek, Schultz reminisced “On and on, past buttes and high ridges, over stretches of level plain, by many a herd of buffalo and antelope, and far in the night arrived at our destination, tired and hungry. We had no thought that all that game we saw was soon to vanish, and that the wide plains we crossed were soon to be dotted with vast herds of the accursed sheep” (Schultz p. 108).

Effect of grazing by cattle and sheep on fire frequency at CMR. Savage and Swetnam (1990) documented fire decline following introduction of sheep into ponderosa pine in the early 1800’s in Navaho country and a similar phenomenon should have occurred at CMR. By 1900 the Montana Bureau of Agriculture reported that in Choteau County, which at the time included what is now western Phillips county “…more than 150,000 head of cattle and more than half a million head of sheep range and fatten on the rich and nutritious grasses that cover the uplands”.

Fergus county shipped more than 4 million pounds of wool in 1900 and the range appeared to be saturated with cattle and sheep. The eastern two thirds of the future Phillips county was still part of Valley county `where the State Bureau of Agriculture (1900, p. 169-170) wrote this account aimed at attracting homesteaders.

“The soil of the valleys is generally a rich black loam that produces abundant crops of grain, grasses, vegetables and root crops. But in a very few years irrigation will certainly make changes for the better all over the county The valleys will be turned into great hay fields and the bench land breaks and bad lands into pastures and range for stock. Malta is next to the largest town in the county for population and business. Here during the summer last past [1899] was shipped 18,000 beef cattle 20,000 mutton sheep 1,000,000 pounds of wool and 175,000 sheep were shorn by two sheep shearing plants.” Around Glasgow “ The surrounding county is settling up with cattle and sheep men.” “Valley county has room for thousands of industrious men and will welcome cordially all who seek a home within her borders. The climate is healthy and vigorous, the resources are abundant and there is no reason why a man should fail who labors industriously and energetically for success.”

This propaganda campaign, though detested by the big ranchers such as James Fergus, led to the storm of homesteaders that flocked in to the green, well watered, and fertile loamy land of Eden from 1905 through 1920.

Fire spread in grazed and overgrazed shortgrass prairie. Wildfire behavior in grazed and overgrazed shortgrass prairie today can be evaluated to give us an idea of worst-case scenario of the range of fire frequency reduction caused by buffalo grazing in the original landscape. In the figure above, the burn patchiness resulted in part from heavy grazing pressure and in part from patches of nearly barren soil and gravel, and yet the fire carried through the landscape and would have kept on going in the absence of suppression. The downwind landscape had better grass cover and the fire, on west winds, would likely have carried as far as the junction of the Missouri and Musselshell Rivers.

Figure 29. August 2007 fire in overgrazed shortgrass prairie, was stopped by firefighters at a road seen on the upper left (Musselshell Trail). This site, just outside the refuge boundary was heavily grazed, as judged by sample density of cow dropping per hectare, and by the large number of cattle and bare ground around a stock pond nearby.

Effects of cattle. Cattle were driven along as a milk supply on trapping expeditions as early as 1833 (Larpenteur 1898) into western Wyoming and down the Yellowstone River, probably the first cattle in Montana. From that time there were cattle in at Fort Union downstream from CMR on the Montana/North Dakota line (Larpenteur 1898) where they was a hayfield and hay was routinely put up for the winter.

The graphs below show numbers of cattle, sheep and horses in Montana from 1867 to the present (supplied by Dan Harrell)

Figure 30a.

Figure 30b.

Figure 30c.

Figure 30d

Figure 30e

Figure 30f.

55 4) TIMBER USE AND FIRE Other than clearing cottonwood bottoms for farming, cutting of fuelwood and use of cottonwood and pine timber for construction of trading posts and other log buildings, the only systematic use of timber in the breaks was cutting by woodhawks for sale to the steamboats on the Missouri. This left some sections in virgin timber.

Woodhawks, during their approximately 28 year era in the CMR area (1862-1890) , cut and sold wood for fueling steamboats. Charles Larpenteur in his 40 years managing trading posts at Fort Union and elsewhere recorded the comings and goings of the steamers. The first steamboat reached the Montana border around 1833 or 1834. In the early days there was only one a year bringing supplies for St Louis to Fort Union and returning downriver with the winter season’s returns of furs and buffalo robes. In 1834 the steamboat Assiniboine made it as far upriver as Fort Union on the Montana-North Dakota line and for a number of years this was only an annual event.

There was still only one steamer, the American Fur Company’s Antelope in 1838 (Larpenteur p. 121), and by 1840 the trip upriver from St. Louis to Fort Union still took two full months with the primitive early equipment. In 1846 Larpenteur described cordeling a keelboat up through CMR to Fort Louis (near Benton). The trip took 70 days but all he had to say about the passage concerned killing 35 deer and 15 elk on the way (p 210).

By June 1861, about 27 years after the first steamboat, multiple boats were traveling upriver and two steamers arrived at Fort Union, the Chippewa and the Spread Eagle. The Chippewa intended to continue up to Poplar River (between Fort Union and Milk River) but was destroyed by a fire started by a drunk.

1862 marks the beginning of the era of woodhawks in the CMR area, as enough boats started coming to make it worthwhile to set up small woodyards at steamer landings to sell fuelwood. In that year Larpenteur recorded two steamers in June to Fort Stewart, the Shreveport & the Emilie (p 288) and in spring, 1863 there were three, the Robert Campbell, the Nellie Rogers and the Shreveport, at least two of which went up through CMR to Fort Benton. When the third encountered low water, goods had to be offloaded at Milk River and taken up to Fort Benton overland by wagons (p.298).

In 1864 the large steamer Benton arrived at Fort Union with 50 tons of Army commissary freight and 150 men. A second steamer, the Yellowstone passed through in June on its way to Fort Benton. The Army also had 5 steamers pass through Fort Union on the way to the Yellowstone River to fight Indians (p. 306).

James Schultz, 1877-1882, commented that in the summers when there was no trading, “The arrival of a steamboat now and then with our mail was the only thing to break the monotony of the days.” In “Floating on the Missouri, Schultz mention woodhawks in the CMR several times. There was a woodyard at the mouth of the Musselshell River and another downstream. The yard at the Musselshell was being operated by William Downs in 1884, when he was killed by a gang (Schultz 1902, p. 72). He also described an earlier fight there between some woodhawks and a band of the Yanktonais Sioux (p. 71).

A second woodyard was located downstream between the Sphinx and Round Butte where Schultz’s friend R.L. McGonnigal cut wood in the 1870s (Schultz p. 117) and it seems likely that there would have been a similar operation at the trading post near Carroll.

Downstream from CMR the fuel for steamboats consisted only of cottonwood, that being the only timber available for hundreds of miles. In CMR the woodhawks switched to pine. By the mid 1860’s all the timber easily accessible near the river had been cut in one or two areas with suitable landings for steamboats. George Croft, a former woodhawk who had a woodyard upstream from CMR at the Coal Banks told Schultz that “In the fall of ’67 we moved down to the mouth of the Judith and started to get out wood for the steamboats there, having cut and sold all there was in the vicinity of the Coal Banks [upstream

56 from the Judith River] (Schultz p.29). “We had six men in our employ cutting pine up in the breaks and in the hills, but one of them was always on the lookout for any sneaking war party, while the rest worked. Nat and I hauled the wood to the river with three yokes of bulls (oxen). We had no horses, and we took turns going after the cattle in the mornings (Schultz p. 28-30).” He noted that they cut the pine with axes.

The woodhawks at CMR, until the last steamboats were replaced by the two railroads around 1890 cut all the timber locally in at least a few spots, primarily around steamboat landings and trading posts. It seems unlikely that they affected the larger landscape and many existing trees only a short distance from the river predate their era by 100 years or more as evidenced by the three fire scar chronologies. There is nothing to suggest that the activities of the few scattered woodhawk operations had any effect on fire frequency.

5) FIRE SCAR CHRONOLOGIES

Composite fire scar chronologies were prepared for three sites: Sand Creek, Soda Creek and Lost Creek. Developing chronologies involves precisely measuring tree ring widths using a stage micrometer on whole sections through trees and identifying fire scars and their dates of occurrence.

Most of the tree sections obtained from the three sites were from ponderosa pine and a few Douglas firs killed in the 2006 wildfires. A few live trees with conspicuous fire scars with evidence of multiple fires recorded in the scar were also taken. A number of sections cut by Hedman in 1983 in the Soda Creek vicinity had been stored at the refuge and were also included. All sections were taken to the Tree Ring Laboratory at Columbia, Missouri for preparation and analysis. Many of the trees were much older than expected, many annual rings being missing because of drought years. Missing rings were detected using a computer program that compares short segments of tree rings from each section with those of sections of other trees. In general drought years, micro-rings that appear in most trees can serve as a marker for that year. A regional master chronology is constructed using a number of trees from the different sites. A computer method is used to match segments containing that marker and others with another segment. The result is that segments with missing rings can be identified. Identifying he missing years is critical for accurately dating the fire years.

Considerations in using fire scar chronologies for determining historical fire frequency. A number of things have to be considered when interpreting the fire frequency obtained. Underestimates may result from several things. Where fuels consist only of pine needles or of light grass barely able to carry fire as often is found in shortgrass prairie, only one or two trees or none at all, of the perhaps twenty stems we have to work with, may be scarred. The data becomes thinner, the farther back in time we go just because many trees have died and there are fewer stems to detect light fires. We might suspect that underestimates are more likely that overestimates in shortgrass prairie because of the light fuels. Other considerations are related to fire frequency: the more frequent the fire the less likely that a tree will be scarred. At the other extreme, the longer the fire interval the more likely that a scar will heal over and stop recording future fires. The longer the fire interval, the more likely that juniper or other fuel may build up beneath a stem increasing the likelihood that it will be killed. This is probably of less consequence in determining true fire frequency because the hotter fires will scar more trees and there should be some left for study. If a very hot fire killed most of the older trees that may leave so few that a number of light fires will be missed.

Sand Creek: evidence for original fire frequency at Sand Creek is ambiguous, partly because of difficulties with being certain that the scars seen were from fire rather than other causes such as porcupines. The site is somewhat fire sheltered by the Missouri River and by steep slopes along Sand Creek as indicated by fire refugial Douglas fir in the ravines. The study site is partially fire sheltered and fire from the south or southwest would have to first cross Sand Creek. Disregarding the questionable scars would give fie

57 intervals anywhere from 22.8 (20th century) to 100+. The site does not represent the higher fire frequency on the upland prairies or for the more fire exposed location at Sand Creek Station.

Figure 31. Fire scar chronology for Sand Creek.

Lost Creek

Figure 32. Chart of historical fires detected at Lost Creek, east central CMR. The oldest stem collected germinated sometime before 1600. For the 406 years of fire history shown here the pattern is of somewhat frequent light fires in grass, interrupted at very long intervals by severe fires that scarred many trees over the whole site. The area sampled was considered to be one fire compartment, that is a unit of the landscape that would tend to burn as a whole .

Suggestion of a higher fire frequency in the pre-horse era. Both the Sand Creek and Lost Creek chronologies have some scarred trees that go back to the pre-horse era. There are too few recorded trees to give a reliable estimate but the two intervals seen at Lost Creek are 8 years and 7 years. Considering that there were only five trees recording, there may have been other fires that would have been picked up if there other trees that old available and it seems likely that fire was occurring more often than 7 years. Similarly at Sand Creek there is a cluster of scars close together between around 1655 and 1720, giving a mean fire interval of about 9 years. That is with only three trees recording, so it is likely that a larger sample would show a fire interval less than 9 years. Sketchy as this is, both sites show a gap in fire interval after the advent of horses around 1730. At Sand Creek there was only one fire in the 140 year period 1720 to 1860 and at Lost Creek there was a gap of 98 years between 1717 and 1815. Once again there are too few trees to trust the data and there were almost certainly other fires that didn’t get recorded but the combined patterns at least hint that there was some decrease in fire after the Blackfeet acquired horses. If true, this could be the result of abandonment of fire as a means to drive bison or to surround them with fire as practiced by other groups.

Soda Creek

Figure 33. Fire scar chronology diagram for Soda Creek.

Buffalo Grazing and Fire – Evidence from the Fire Scar Record In each of the three fire scar chronology sites the recording trees with fire scars after 1853 are numerous enough to give a reasonable picture of past fire history. The fire intervals in the table below are the average between the three sites for the intervals shown. Before 1882 the mean fire interval at the three sites was 13.3 years. The next interval before that was 14.7 years. Then a remarkable thing happened. The winter of 1881-1882 was the last year of the buffalo trade. The last great herd was gathered on the vast plain on the south side of CMR, bounded by the Judith Mountains on the south, the Missouri River on the north and the Musselshell River on the east. The native peoples were largely unaware that these were the last buffalo. As bison had been extirpated in the surrounding lands, the tribes that followed the buffalo surrounded the remaining herd. This was primarily the land of the Blackfeet but there were Crow to the Southeast, Cree to the north, Assiniboine to the east and Flatheads to the west. As the fall and winter hunting season progressed these tribes collapsed in upon the remaining buffalo, slaughtering all except for a small herd of a dozen or so that took refuge in the badlands of Garfield County near the Missiouri River.

For the first time in thousands of years the spring of 1882 saw no buffalo on the prairies. And then, amazingly the mean fire interval among the three sites dropped to 3.3 years. With no grazers the grasses would have abundant, the fuels continuous and fire was apparently able to spread freely. What this shows is that fire frequency went up immediately upon extirpation of the buffalo. The implication of the numbers is that without buffalo fire frequency would have been four times higher in shortgrass prairie. Cattle and sheep were already pouring in, with James Fergus’s 100 head of cattle and horses he brought from Helena to the headwaters of Armells Creek in 1880 and the several thousand sheep his brother brought to the east side of the Judiths shortly after. Their numbers increased by tens of thousands in the next few years so that the next fire interval returned to 12 years, approaching that of the buffalo era.

THE BISON EXTIRPATION EVENT ~1853 ~1868 1882- ~1885-1897 ~1897- ~1917-1946 -1868 -1881 1885 1917 Bison Bison ~No Sheep & Sheep & Heavy grazing Grazers cattle cattle & fire expansion excess suppression Mean Fire 14.7 13.3 3.3 12 19 29.3 Interval (3 sites)

Table 6. Fire return intervals, averaged between the three sites at CMR

Figure 34. Cattle, sheep and horses on open range in Montana 1867-1886. Note that the winter of 1881- 1882 is an inflection point. Livestock numbers began to rise dramatically from that time with cattle and sheep reaching over a million each six years later in 1886.

Figure 35. Livestock numbers in Montana from 1887-1910.

The numbers grew steadily over the next fourteen years reaching a peak in 1903. This level apparently overshot the carrying capacity of the range, as happened in other places in the west upon introduction of livestock, followed by a dieoff of about three-quarters of a million cattle in the next two years. As noted above the number of livestock in the decade 1900-1910 averaged 5.2 million cattle, more than double the number today, and 1.1 million sheep. With this horde consuming the grass fuel, the next fire interval dropped to 19 years and the one after that was 27 years. The implication is that saturation of the range with cattle and sheep reduced fire frequency to less than half its frequency under a natural bison grazing regime. And as we have seen the infilling of the woodland landscape with juniper and trees, producing our m oder fuel loads, began in 1900 and continued for the next forty years.

6) EVIDENCE FROM TREE DEMOGRAPHY PLOTS

Cross-sections of all woody stems were collected from plots 10 meters by 50 meters in stands killed by the July 15, 2006 fires.

Lost Creek. Figure 36 below shows the pattern of infill in a plot at Lost Creek.

62 RECRUITMENT

600 400 DIAM 200 0 DAIMETER (mm) 1880 1900 1920 1940 1960 1980 2000 2020 YEAR

Figure 36. LOST CREEK – History of recruitment in a ponderosa pine stand killed by a lightning fire on 7/15/06

Figure 37. 100% mortality in a ponderosa pine stand at Lost Creek killed by fire. The fire, driven by cold front winds, came from the northwest, up the draw seen below on the left, generating intense heat and killing all stems. What can’t be seen is whatever shrub density occupied the understory before the fire

In the demography plot above there were no trees older than about 1900, the time of first saturation of the landscape by sheep and cattle. In the absence of fire, the stand continued to fill in during the 45 year period 1900-1945. After that time the stand had a closed canopy, was fully stocked and too shady for recruitment of additional saplings. The amount of shrub cover is unknown as juniper can be completely consumed by fire leaving no trace as illustrated by the before and after pictures at Soda Creek.

CK CREEK

300

250 200

150 DIAM_mm

100

0 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 Figure 38. C.K. CREEK – demography of a ponderosa pine stand on the north side of the Missouri River. Diameters are in mm.

CK Creek. demography of a living ponderosa stand on the upper rim on the east side of CK Creek shows a similar pattern to that of Lost Creek with the exception that there are five trees that germinated before 1900. These were near the lower end of the plot in a more fire-sheltered location. As at Lost Creek, after 1900 there was a pulse of new recruitment that continued until the stand was fully stocked. One difference between this plot and that at Lost Creek is that there was a heavy buildup of juniper under the ponderosa at CK Creek (juniper stems not sampled).

64 Figure 39. C.K. Creek, Phillips County. The 10 x 50 meter demography plot was done in the ponderosa stand at top center just below the upland prairie flat extending to the north beyond the rim of the creek. Tree cross-sections for a third demography plot were recently collected from Soda Creek but have not yet been processed.

7) EVIDENCE FROM HISTORICAL PHOTOS AND OTHER FIRE HISTORY

Lewis and Clark mentioned just one fire in CMR, in a cottonwood bottom, that nearly caused a tree to fall on one of the tents. James Willard Schultz traded furs from 1877 through 1904 all up and down the CMR and also explored in the surrounding lands from the Musselshell to the Judith and Marias Rivers. Until his death in 1947 he recorded Blackfeet Indian history and published books and stories about his experiences with them beginning in 1877. He was a keen observer of wildlife, geologic process and all aspects of nature and yet in all his writings I noticed only a single mention of fire. That happened to be in a cottonwood bottom near Hornet Island, now inundated by Fort Peck Lake. This is to the east after rounding U.L. Bend going downstream from the mouth of the Musselshell River. Schultz noted “The next bend below the island on the north side is called Horseshoe Point [also inundated] and is about a mile square.” “Except for a narrow belt of green timber fringing the shore, the whole flat was a tangle of burned and fallen trees, and grown up with the thickest, tallest rose brush I ever saw. No doubt it harbored many deer, but I didn’t want one of them bad enough to venture into the thicket” (Shultz 1901 p. 87). This was just a little above the mouth of Lost Creek and in the most lightning-ignition prone center of CMR. As the fire-killed trees were down, the fire must have occurred some ten or so years before, around 1890.

Fire prone and fire sheltered sites can occur in close proximity. Below Devils Creek the river enters a narrow canyon with rugged, steep walls. Here, in this fire-sheltered stretch, Schultz remarked on the presence of Douglas fir. “From Herman’s ranch to the mouth of Seven Blackfeet Creek, the southern rim of the valley is one continuous cliff of sandstone, pierced by walled coulees, capped with a lovely fringe of green timber. And on the slopes, below the frowning walls, stretches the heaviest growth of pines and firs of any we had seen.”

Figure 40. Grassland on way down to Hell Creek, 1902. I spent several hours looking for this site in hope of taking a matching photo but there is little to go on except for the

65 shape of the ridgeline when viewed from the correct angle. Although with enough effort it should still be possible to locate this site, I was unable to find any areas this grassy without juniper or pine around the upper rim of the Hell Creek drainage. This suggests sufficiently high fire frequency to prevent colonization by woody species before 1902. Several buffalo skulls can be seen in the foreground and in the distance. L..A. Huffman photo “Buffalo trails at Hell Creek” courtesy of the Montana Historical Society Library.

66 Figure 41. One of 14 sets of repeat photos by Shantz, Phllips and Kay showing vegetation change in the Judith Mountains (Klement, KD, RK Heitschmidt and CE Kay. 2001).

In Figure 41 the first photo in 1917 was taken 17 years after the landscape became saturated with sheep and cattle. The fire scar chronologies show that 1900 marked the beginning of infill of such stands but establishment of a new crop of saplings is a 40 year process in this climate. After 17 years the new saplings established to date would still be small so it seems likely that the amount of tree cover seen in the first photo is close to the natural stem density in balance with the pre_European fire regime. In all the photos the trees are densest on the north and northeast-facing slopes. before 1900, with fairly frequent fire, on the order of 3-10 years in the grassy Judith Mountains, there would be no time for juniper or other woody fuel to accumulate so old ponderosa pines would be unharmed by the light grass fires. Recruitment of a new stem here or there would be a rare event because most new seedlings and saplings would be eliminated by fire. Only an occasional stem, perhaps in a slightly fire sheltered microsite such as behind a boulder or in a patch of bare soil would reach a large enough size and build enough bark to resist fire. This explains the thin pattern of trees seen in the 1917 and other historical photos.

Despite the superabundance of livestock, severe burning conditions could still produce a fire in this era. In 1910 “…it was extremely dry and fires burned from the Judith to the Musselshell, blackening thousands of acres” (Willmore, 1987, p. 3). The fire movement described from west to east sounds like a post-cold front fire with winds from the west or northwest required to drive fires in this direction. As seen in Figure 42 below, even overgrazed areas may still carry fire when associated with strong wind.

Figure 42. August 2007 fire along Musselshell Trail was stopped by fire fighters along the road at upper right. Despite overgrazing by cattle the fire carried through the landscape leaving unburned patches. I watched this small storm, spawned off the Judiths on southwest winds at sunset the night before. The fire, if not suppressed, would likely have made its way to the Missouri and Musselshell Rivers several miles to the east.

Figures 43 and 44 below are a matching pair taken 127 years apart. The scene is the mouth of Cow Creek, upstream from CMR, facing east. There is a small Army encampment on the flat west of the creek mouth which can be seen behind and slightly to the left of the steamboat. The line of cottonwoods in the distance is Bull Creek. Taken in 1880, this is definitely pre-settlement, the year before the great buffalo slaughter on the plains just to the south. The Blackfeet were still hunting buffalo on those plains at the time the photo was taken. The camp flat, about eight feet above the river level was too high and dry for establishment of cottonwoods although it likely supported them in earlier times before the river cut down to its present level, lowering the water table. The island in the river in the foreground may be new or more likely the water level was higher at time of the 1882 photo. Larpenteur and others mention that the steamboats only came up the river to Fort Benton on the spring rise. At other times of year goods had to be offloaded at places like Milk River or Carroll and freighted overland by ox teams. The recent photo shows a substantial increase in cover of ponderosa pine over that of 1880. Although the hills look barren in1880, close examination of the original photo showed a few trees. To compare vegetation on the slightly fuzzy 1880 photo, look at the small bump on the highest hill at left center and then notice that that feature has been obscured by trees in the modern photo.

Figure 43. Photo taken by F.J. Haynes, 1880 at the mouth of Cow Creek

Figure 44. Matching photo taken August 2007. The sagebrush, almost entirely silver sage, as seen in a close-up photo in the army camp, has been trampled back to the margins of the flat by cattle, which were present but out of sight at time of the modern photo.

Figure 45. Close-up of the portion of the ridge seen on the right side in the 1880 photo showing the modern extent of ponderosa pine and juniper.

Figure 46, also taken at Cow Creek 1880, looking north.

Figure 47. Comparison with Figure 46 shows an increase in woody cover on the near and far hills in the background.

70 8) MAPPING PRE-EUROPEAN FIRE FREQUENCY IN SHORTGRASS PRAIRIE

This section explains how the GIS fire frequency map was made. The process involves a number of steps and synthesizes whatever kinds of information are available. The mapping also takes into account the principles around how fire moves in a natural landscape and the factors that affect both frequency and fire severity.

Table 7. Some kinds of evidence used in mapping original fire regimes

BIOTIC EVIDENCE . Fire frequency indicator species (proxies for fire frequency) . Fire frequency indicator communities (proxies for fire frequency) . Reduction in fire frequency by native grazers . Patterns of plant succession and fuel buildup with reduction in fire LANDSCAPE AND ENVIRONMENT FACTORS . Original fire compartment size . Presence of fire barriers and fire filters . Effects of soil productivity on fire behavior (density & continuity of grass) . Fixed lightning generators, lightning strike density and ignition records

HISTORICAL EVIDENCE . Fire scar chronologies . General Land Office Survey witness trees . Vegetation types on early metes and bounds survey maps . Herbarium records of fire frequency indicator species . Historical references to use of fire by Native Americans . Vegetation types on old aerial photos.

Table 8. NINE KINDS OF FIRE HISTORY EVIDENCE AVAILABLE AT CMR 1. The three fire-scar chronologies 2. Area lightning ignition patterns 3. 27 years of refuge records of ignitions and area burned 4. Fire frequency indicator plant communities such as those with Juniperus horizontalis or Douglas fir 5. Likely fire frequency indicator species such as Echinacea angustifolia and Orthocarpus luteus 6. 1/10 hectare study plots 7. Tree demography studies 8. Historical photos and matching photos taken later 9. Historical descriptions by Lewis & Clark 1805-1806, Larpenteur 1833-1872, Schultz 1877-1901 and others

A first step when doing the field work is to recognize that you are in a fire landscape and then to always to be aware of your position in that landscape. Useful questions may be, how fire sheltered or how fire

71 exposed is this spot? What are the pathways fire can follow to reach it? How do the pathways change if the winds are from the southwest versus cold front winds from the west, northwest and north?

FIRE EXPOSED VERSUS FIRE SHELTERED SITES

Some examples of the most fire exposed (top) to fire sheltered (bottom) sites at CMR: 1 Lost Creek ponderosa pine parkland 2 Upland prairies with loamy and fine sandy loam soils (produce more lush grass fuel) 3 Upland prairies in general 4 Upper slopes and slope shoulders in the breaks 5 Lower half of steepest-sided ravines in the pine breaks 6 Missouri River cottonwood flats 7 Beaver wetlands 8 Douglas fir pockets in deep ravines with no access by fire 9 Badlands of eastern and northern Garfield County 10 Rolling gumbo barrens of Valley County between Willow Creek and Fort Peck Lake

Some Steps In Simultaneous Mapping Both Pre-European Settlement Fire Frequency and Vegetation First Approximation: 1. Preparing GIS base map: • soils • topography (LIDAR-based in flat landscapes) • streams • roads 2. Prepare a first approximation vegetation map based on field plots 3. Prepare a relative fire frequency map 4. Approximation of presettlement veg. community types based on 1/10 ha plots Second Approximation 5. 2nd approximation vegetation map (adjusted for what the fire frequency map tells you) 6. Continue vegetation plots and begin to collect historical accounts and photos. 7. Collect GLO surveys or early land grant survey plats 8. Rank species and interpret fire relations of witness trees. Repeat for plant communities 9. Refine both the presettlement vegetation and fire frequency maps in the light of historical records 10. Examine historical literature (including accounts of early travelers) Third Approximation 11. Assign real numbers to fire frequency map based on fire scar chronologies 12 Assign numbers to fire frequency map based on proxies for fire frequency such as fire frequency indicator species and indicator communities 13 3rd approximation vegetation map (based on final fire frequency) 14 Assign acceptable species to each category of original vegetation and fire frequency 15 Test final results to determine accuracy of classifications 16 Make final revisions to maps

Making a Relative Fire Frequency Map

The first step in constructing a fire frequency map is to make a relative fire frequency map. You can make a relative fire frequency map for any landscape even if you know little about the actual fire frequency numbers. First locate the most fire exposed point in the landscape and the most fire sheltered point. For CMR these would be the grassy prairies on the upper rim of the breaks on the soils that are most productive of dense, continuous grass cover, and the oldest Douglas fir stands in the deepest, most fire sheltered ravines along the Missouri River. At CMR, using the three fire scar chronologies, we were able

72 to get a presettlement fire frequency of around 8 years for the most fire exposed areas, while there are Douglas fir stands over 300 years old in the deep ravines. That gives us the end points of the fire frequency gradient. Next comes interpolation of fire frequency for the intervening parts of the landscape. For this it is useful to chop the fire frequency gradient into classes to make it workable. Table 10 below shows the nine classes developed for CMR. Then it is necessary to break the landscape into fire compartments to serve as mapping units.

Drawing Fire Compartment Boundaries. This process must be done by hand as there is as yet no computer method available. Fire spread models such as FARSITE and fire intensity models such as FLAMMAP can provide starting points but the elements contributing to fire frequency are too complex to fit any existing model. The necessary materials and elements of a fire frequency map on GIS include, layers for topography, soils, hydrology and roads. Other requirements may be needed such as a table of soil productivities in tons per acre (as a way to estimate grass fuel quantities sufficient to provide fire spread on each soil type), a map of prairie dog habitats (as fire filter supplements to bare soils and badlands as elements reducing fire spread), a knowledge of the dominant ignition sources and patterns of fire movement, as just discussed above, as well as historical information on numbers of buffalo and livestock.

Fire compartment definition: a fire compartment is a unit of the landscape with no internal firebreaks and nearly continuous fuels, such as grass or pine needle litter, so that an ignition in one part would be likely to burn the whole unless there were a change in weather or fuel moisture.

In Figure 48 below fire compartments are numbered in orange (the letters in yellow are the fire frequency classes). Boundaries are drawn for fire compartments using rivers, streams and coulees, without consideration for whether they represent strong or weak firebreaks. The point is to divide the landscape into small units that can be considered individually for assigning fire frequency.

Where there is an obvious local fire frequency gradient without a stream of other feature to use as a boundary it is necessary to draw and digitize a boundary by hand to break the gradient into separate units. For example, compartment 51 is on a peninsula formed by Crooked Creek and the Missouri River and is more sheltered by those features than the land just to the northwest, so the gradient was broken by a hand- drawn line, following low points in the local topography, to form compartments 50 and 51. In some cases a vegetation type may be assigned its own fire frequency class, for example, beaver wetlands in class G, shown in green below. In other cases, barren soils can be used as boundaries as in the gray area in the lower right corner labeled FF. Finally the compartments are numbered. In all there were 134 fire compartments so delineated at CMR.

Figure 48. Sample portion of the completed fire frequency map. The next step is assigning tentative fire frequency classes to each of the 134 fire compartments, still without worrying about the fire frequency numbers these classes are going to represent.

Assigning Fire Frequency Classes to Fire Compartments

It is convenient to chop the fire frequency gradient into about eight or nine fire frequency classes that will form the map units. Later actual fire frequency numbers can be assigned to those classes for which there are data, such as from fire scar chronologies and fire frequency indicator species. The frequency for compartments for which there is no data can be interpolated from the others. For CMR it seemed appropriate to have six classes representing a gradient for fire frequency with A being highest and F lowest. In addition three more special classes were added, G to represent several bottomland vegetation types with a highly variable fire frequency, FB to represent portions of the landscape acting as firebreaks and FF for portions functioning as fire filters.

Map colors were chosen for classes A through F with the lightest to represent the highest fire frequency to suggest the more open, sunny nature of frequently burned areas, and compartments with lower fire frequency shown progressively darker to correspond with the dark, shady wooded communities like Douglas fir or old growth cottonwood. Class G represents a special group of vegetation types with a low fire frequency that is variable because of much local variation in access by fire from more fire frequent

74 uplands. Class FB (for fire break): the darkest gray color, was reserved for essentially fire-free types and a lighter shade of gray was used for Class FF (for fire filter).

On the draft GIS map all the compartments were numbered. Then for each compartment all factors influencing fire frequency were evaluated for assigning the compartment to a relative fire frequency class.

Assessing landscape factors influencing fire frequency

Fire RFF Side Fans Wind FC Prod Prod Access PD PD Isol Comp 1 Size In Up- In Up- No. wind wind 200 D 2 1 1 3 2 3 3 0 13 3

Table 9. Landscape factors influencing fire frequency. In one step each of the 134 fire compartments was evaluated according to eleven parameters affecting fire frequency and each compartments were assigned to one of nine fire frequency classes along the fire frequency gradient. Scoring landscape influence on fire frequency was done as follows. Parameters were scored so that the lowest number correlates with the highest fire frequency, higher numbers correlate with greater degree of shelter from fire.

Key to Table above: Fire comp no - Fire compartment number RFF - Relative fire frequency (becomes the fire frequency class A through F when fire frequency numbers are assigned) Side - 1 = south side, 2 = north side of the Missouri River, which being downwind from the most frequent ignition sources should have a slightly lower fire frequency than the south side. Fans - represents the landscape position in the areas of overlap of the lightning ignition fans shown in Figure 18. 1 means that the fire compartment is in the highest ignition density area in the center of the refuge where all four fans overlap (from the Judith and Snowy Mountains, the Little Rocky Mountains and the Bearpaws), 2 indicates 3 fans, 3 indicates two, 4 indicates one and 5 indicates none. Each can also be scored by distance from the center of the axis: 1 = near, 2 = away from axis, 3 = far. Wind1 – primary wind direction driving fire spread for fires that affect this compartment. for example, compartments in the vicinity of the Sand Creek Station and west of Highway 191 receive most of their ignitions from small, short-lived storms ignited by orographic uplift in the Snowy or Judith Mountains, associated with southwest winds. Wind2 – secondary wind direction, WNW for rare dry cold front ignitions from the Bearpaw Mountains in the Sand Creek example. FC size – size of the individual fire compartment: the large the size the higher the fire frequency, scored 1 = large, 2 = medium, 3 = small. Prod in – Productivity or fuel continuity within the compartment. NRCS figures for pounds per acre of range grass productivity were used for prairies and presence or absence of bare soil or rock were used for wooded areas where pine needles and thin grass carry fire. The more productive and continuous the fine fuel the higher the fire frequency. 1 = high fuel productivity and continuity, 3 = low Prod upwind – If there was a fire path connecting the particular fire compartment to ignition sources upwind, it was scored 1 = higher fire frequency, 2 = medium, 3 = low. The longer the fire path the higher the fire frequency because the additional area increases the effective fire compartment size and because the path leads closer to the lightning source where ignition frequency should be higher. A fire path is a continuous sequence of productive soils with better than average grass cover, forming continuous fuel unbroken by any rock outcrops, barren areas or any other firebreaks or fire filters. The path also may intercept fires moving through the upwind landscape that might otherwise have not reached the fire compartment. Bringing in fire from outside augments the fire frequency within the compartment.

75 Access 1-3 refers to fire access issues such as whether the compartment lies transverse to or parallel with the prevailing winds. An elongated compartment transverse to the prevailing winds would have more exposure to intercepting a moving fire than one that is parallel to fire winds. Also, a fire compartment with access to fire only through a bottleneck at one end would have lower fire frequency than one with wider access or access in the middle. PD in - scored 1–2. A compartment that contains prairie dog towns within could have reduced fire frequency: 1= no existing or historic prairie dog towns as predicted by the prairie dog habitat layer, 2 = prairie dogs, 3 = large percent of compartment has soils, moisture, slope and landforms highly preferred by prairie dogs. PD upwind – scored 1 – 3 = prairie dog towns now or historically existing upwind from the fire compartment under consideration. Prairie dogs towns upwind could act as fire filters and reduce fire frequency downwind. 1 = none, 2 = some, 3 = abundant prairie dog towns or existence of preferred soils and habitat for prairie dogs (based on the GIS layer for that purpose). Isol - Isolation of compartment from pathways for fire flow: ignition primarily internal (from lightning strikes inside the compartment) 1 = not isolated, 2 moderately isolated, 3 strongly isolated from fire approach by the river, rugged topography, rock outcrops or badlands.

The scores obtained above can be used to assign and refine the relative fire frequency classes A through F.

Assigning fire frequency numbers to fire frequency classes

The next iteration involves assigning actual fire frequency numbers to the nine classes of fire. The anchor points for assigning actual fire frequency were the fire history windows obtained at three points where fire scar chronologies were prepared from sections of ponderosa pines and Douglas fir.

1) Fire frequencies obtained in the fire scar chronologies are assigned to their respective fire compartments. 2) The same number and fire frequency range is assigned to other compartments in the same fire frequency class. 3) Adjacent compartments are assigned different fire frequencies if they were in higher or lower relative fire frequency classes. 4) Slight adjustments are made for any fire frequency indicator plant communities such as those with Juniperus horizontalis or Douglas fir and likely fire frequency indicator species such as Echinacea angustifolia and Orthocarpus luteus 5) Adjustments may be made based on evidence from 1/10 hectare study plots, tree demography studies, historical photos and historical descriptions by Lewis & Clark 1805-1806, Larpenteur 1833-1872, Schultz 1877-1901 and others

Then as a final step, the last adjustments are made based on the pattern of actual records of fire frequency for the 27 years for which data are available.

Figure 49. Ranking vegetation types by fire interval. Juniperus horizontalis-J. communis-J. scopulorum, a fire frequency indicator community for Fire Class D, 12-35 years

Fire Frequency Classes – Variation Between Original Fire Frequency Before and After Introduction of the Horse to North America by the Spanish

Table 10 below summarizes the fire frequency classes used for mapping fire frequencies at CMR for the pre-European settlement period 1730-1882. This is for the period during which the Blackfeet and other tribes used horses for hunting the buffalo. A second version covers the pre-horse era before 1730 when there was likely more use of fire by Native Americans, for which there is some support from the fire scar chronologies.

Table 10. Fire Frequency Classes at CMR 1730-1882 after introduction of the Horse by the Spanish Fire Mean Historic Range of Variation Frequency Fire (90% of fires, estimated from fire Class Interval scar chronologies and other (years) evidence) (years) A 8 4-14 B 10 6-20 C 14 8-30 D 22 12-35 E 40 25-70

77 F 100 70-700: compartments contain some highly fire-sheltered sites G Infrequent, Fire frequency depends upon location Variable in the landscape and accessibility to fire from upland sources FB None Firebreaks: barren gumbo or shale, lacking enough fuel to carry fire FF Limited fire Fire Filters: patch mosaics of spread vegetation with substantial areas of bare soil. Not all fires make it through these zones, reducing fire frequency downwind. Fire compartment numbers

Descriptions of the Nine Fire Frequency Classes:

CLASS A, 8 years (4-14) yrs: This, the highest fire frequency class, was assigned only to portions of the Lost Creek uplands and some other upland prairies and upper slope shoulder wooded stands on the south side of the river in the central section of the refuge. These are the sites with the highest exposure to lightning and accessible by fire from broad expanses of upland prairie. These few areas lie downwind from prairies on the better soils, more productive of grass fuels and lacking badlands or zones of poor soils that would act as fire filters. With continuous grass fuels, ignitions potentially can occur many miles away toward the Judiths or the Snowy Mountains. The longer the fire path with good fuels, the larger the fire compartment size and the broader the ignition area is in the upwind landscape, all of which results in higher fire frequency. Some of these class A sites are on isolated polygons of a particular soil.

How can a patch have a higher fire frequency than the surrounding landscape? The level to gently rolling sandy loam soils support richer grass cover than most of the rest of CMR with exception of the moist bottomland soils along the river. These grassy ponderosa pine parklands are mesa-like erosion remnants of the extensive prairie tableland just to the south. Mickey Butte and Brandon Butte are small mesas only 14 meters higher than the ponderosa uplands of the Lost Creek area which appear to be slightly erosion-lowered remnants of the same original tableland. Under mild burning conditions, fire in eroded lands will travel a certain distance before reaching a rock outcrop, barren patch or deep ravine and go out. The average patch size of burns is smaller and will give the eroded area on average a lower fire frequency than would obtain in a less dissected area. Within such a landscape, an area of rich, continuous grass cover, such as that at Lost Creek pine savannas will usually burn completely, whether or not it is burning upwind or downwind, it matters little where the ignition point is. So if a fire in the more rugged lands touches the upland savanna on one side it will burn the whole and if another fire touches the other side it will burn the whole. The result it that if natural fire frequency in the lands on either side averages 16 years, the fire frequency in the savanna could average 8 years. This is a simplistic model but illustrates one of several, not immediately intuitive, landscape principles that can produce much higher and much lower fire frequency in parts of the landscape that may only be a hundred meters apart and receive exactly the same rate of lightning ignition!

For another example, a long grassy body, especially if aligned with the prevailing fire winds will have a higher fire frequency than the surrounding vegetation since any ignition that touches an edge will take off and run for miles, overlapping previous fires. Selected Vegetation Types: 2.1 (as mapped in a few selected patches) 3.1 (as mapped in a few selected patches)

78 CLASS B, 10 (6-20) yrs: The majority of upland prairies fall into this class in parts of the landscape such as the upland prairie around the margins and the larger peninsulas interpenetrating the Missouri River breaks. Selected Vegetation Types: 2.1 Some isolated bodies ( as delineated)) 2.2 Much of the dry prairie 2.4 Some bottomland flats, playas, saline flats, alkaline big sagebrush flats 3.1 (as mapped in selected patches) 3.3 (as mapped in selected patches)

CLASS C, 14 (8-30) yrs: This is also a frequent fire class but may be slightly fire sheltered in slightly lower areas or on narrower peninsulas where, because of the intervening ravines, the sites may be slightly less well connected to the broad upland prairies. Selected Vegetation Types: 2.3 A few parts of the dry prairie-juniper shrubland mosaic 2.5 Mesic and dry-mesic bottomland needlegrass-wheatgrass prairie (& shrubland) 3.2 Ponderosa pine-juniper-dry prairie mosaic 5.1 Dry drains and coulees

CLASS D, 22 (12-35) yrs: Sites more sheltered by topography, the Missouri River or its major tributaries or by patches of barrens or badland soils Selected Vegetation Types: 2.3 Most of the dry prairie-juniper shrubland mosaic 3.2 Ponderosa pine-juniper-dry prairie mosaic 3.3 (certain slopes, as individually mapped)

CLASS E, 40 (25-70) yrs: Moderately to strongly fire sheltered by topography, the Missouri River or its major tributaries or well isolated from fire by patches of barrens or badland soils. Selected Vegetation Types: 3.3 (certain slopes, as individually mapped)

CLASS F, (100) 70-700 yrs Deep ravines and other places strongly fire sheltered by topography, the Missouri River or its major tributaries or zones of rock outcrops or barren soils. The very oldest stands of Douglas fir occur in small, deep pockets Selected Vegetation Types: 3.3 Ponderosa pine slopes and ravines. The most fire-sheltered Douglas fir ravines Small upland pockets with single trees or small clumps of ponderosa pine on barren gumbo soils or in bands along rapidly upper slope shoulders with too little fuel continuity to carry fire (included in 2.1 and 2.4 but not mapped separately).

CLASS G, variable fire frequency These are bottomland areas where fire frequency is highly variable. In the case of cottonwood flats, a steep, unvegetated cut bank may prevent fire from accessing one flat while another may be connected to a gentle slope with a continuous fuel path leading to a fire source on the uplands. Selected Vegetation Types: 7.1 Cottonwood flats 4.1 Toe slope shrublands (a minor map type.

79 4.2 High river & stream terraces 6.2 Wooded ravines

CLASS FB, firebreaks with little or no fire. Examples include almost completely barren, rapidly eroding gray shale along the Missouri River, often with narrow bands of greasewood and saltbush along horizontal seams in strata where water is available, or where soil texture permits root penetration, and zone of rolling gray shale with ephemeral vegetation such as Atriplex dioica which is too sparse to carry fire. Selected Vegetation Types: 1.1 “Rock” outcrops and poorly consolidated geologic sediments 1.4 The four true badlands soil series. 6.1 and 6.3 beaver wetlands 7.2 The wettest bottomland type along the river, essentially fire free. They tend to act as firebreaks or fire filters of various strengths.

CLASS FF, fire filters. Soils with poor productivity, resulting in grass/forb stands too thin or with fuels too discontinuous to reliably carry fire so that not all fires make it through the sites resulting in lower average fire frequency on the downwind side. Selected Vegetation Types: 1.2 Rock outcrop-vegetation mosaic (mostly on the Bearpaw shale in western half of CMR) 1.3 Badlands soils 1.4 Dwarf Artemisia tridentata flats-dry ridgetop prairie mosaic

Skewness in the fire frequency distributions. The fire frequency curve for any class tends to be skewed to the left. A curve for a class with a three year mean fire interval has a short tail on the left because the only options are one or two year intervals while on the right side there can be a longer tail because there is the possibility of an occasional longer interval of five, eight or twelve years between fires. This is why the estimated Historic Range of Variation in the tables in not symmetrically distributed around the mean fire interval.

Fire Frequency in the Pre-Horse Era

Table 11 below is adjusted to give some idea of fire frequencies that might have prevailed at CMR before 1730. Without horses, native American settlements were more closely tied to major streams with permanent water such as the Missouri river, so, if they used fire in the same ways as the Mandans did before they obtained horses we could expect higher fire frequency at CMR since it follows the river. The two pre-horse fire intervals we have for Lost Creek, with fire scars in 1702, 1710 and 1717 give intervals of 8 and 7 years, or an average of 7.5 years. At Soda Creek there was only one scar before 1730 so we can get no fire frequency interval for that time at that site. At Sand Creek there are scars in 1653, 1664, 1675, 1679, 1692, 1705, 1716 and 1718. If all these are fire scars that gives a mean fire interval of 9.3 years for that site (range 2 to 14 years). Each site has varying degrees of shelter from fire. The Lost Creek site has one ravine between it and the fire-exposed upland prairies to the south, the major source of fire at this location. Sand Creek is a little more sheltered, by Sand Creek to the south and the Missouri River just to the north so a lower fire frequency could be expected there.

For any particular spot at CMR the frequency of fire from lightning ignitions is tied to the fixed pattern of lighting and to the fixed pattern of firebreaks and fire filters, so is predictable and represented by the GIS map of fire frequency. Ignitions by Native Americans should be higher in the desirable locations for villages in the landscape, first with more fire near the village used to attract buffalo to the new grass after fire and second to the use of fire to surround or drive buffalo within foot travel distance from the village. The Mandans did annual burning of the grasslands near the villages. In the mixed grass and tallgrass prairies of their territory, fires could have been expected to spread far. The increase in fire frequency augmented by Native American burning is harder to predict in shortgrass prairie and the pine breaks but

80 ignitions by local peoples should contribute a little increase in fire in the general landscape even though the distribution of fire filter soils makes it unlikely that any one fire would burn very large areas. The table below uses the somewhat skimpy pre-1730 fire scar data to make a best approximation table of fire frequency at CMR before the use of horses.

In making prescriptions for fire, whether to use the immediately pre-European settlement table above or the higher fire frequency pre-horse table below is a management decision. Since the horse was a Spanish introduction, using the pre-horse estimates would certainly be justified since that had been the natural fire regime for thousands of years. Since fires were more frequent they would more often have been low intensity fires in grass and pine needle litter, with only a hot spot here and there.

Table 11. Fire Frequency Classes at CMR Before Introduction of the Horse Around 1730 Fire Mean Historic Range of Variation Frequency Fire (90% of fires, estimated from fire Class Interval scar chronologies and other (years) evidence) (years) A 6 2-12 B 8 4-15 C 12 8-20 D 20 12-25 E 35 25-70 F 100 50-700: compartments contain some highly fire-sheltered sites, rare, random ignitions in pockets G Infrequent, Fire frequency depends upon location Variable in the bottomland landscape and accessibility to fire from upland sources FB None Firebreaks: barren gumbo or shale, lacking enough fuel to carry fire FF Limited fire Fire Filters: patch mosaics of spread vegetation with substantial areas of bare soil. Not all fires make it through these zones, reducing fire frequency downwind.

Considerations for using these tables for prescribed fire. The individual compartments are rather broad and most contain certain areas with slightly more fire sheltered sites and slightly more fire exposed vegetation. It therefore becomes the responsibility of the person in the field to adjust fire frequency based on the degree of shelter they see in the portion being burned. This assumes that the whole compartment will not likely be burned. In nature, any particular fire moving in on, say, the prevailing southwest wind would likely burn most of the more fire-exposed portions and might or might not burn all of the lower portions or those on steep north-facing slopes. If a situation arises where the whole compartment can be burned as a unit, the best way to simulate natural fire is to ignite in the most likely upwind site from which fire would approach naturally and let it burn what it would in the natural process. Since this involves large head fires, the kind that burn the most land in nature, but pose safety or containment problems, most prescribed burns are not likely to be able to completely simulate the natural process.

9) CONCLUSIONS: Results, Products and the Answer to the Question “Is this natural?”

On June 15, 2006 we likely saw the worst that fire can do at CMR. It had been 24 years (2006-1982) since the last severe fire at Lost Creek, with “severe” meaning a fire that scarred multiple trees. The 1982 fire appears to have been more local and did not appear in the chronologies for Soda Creek or Sand Creek. Judging from the pattern of scarred trees it appears that that fire occurred on southwest winds, the more common wind direction while the fires of July 2006 occurred on stronger, post cold front winds. The sum of the results give us a picture of whether or not the severity of these fires was within the natural range.

SUMMARY OF RESULTS

 A fine scale map of pre-European fire frequency was produced (GIS map 1)  A fine scale map of pre-European vegetation and fuel types was produced (GIS map 2)  Three fire scar chronologies (Sand Creek, Soda Creek and Lost Creek) provided anchor points of known fire history on the relative fire frequency map from which we were able to extrapolate to other compartments  Species that are found only in the historically frequent fire areas can be used as fire frequency indicator species. examples are Echinacea angustifolia and Orthocarpus luteus. There may be different indicator species on different soil types.  Fire frequency indicator communities were defined, such as those with Juniperus horizontalis  The fire scar chronologies documented the impact of bison grazing on presettlement fire frequency in shortgrass prairie (fire frequency would have been four times higher without buffalo)  They also documented that the replacement of bison with cattle and sheep doubled the of length of fire intervals, beginning in the 20th century (resulting in half as much fire as there was in the time of the buffalo)

Table 12. The principal products and results of the mapping project.

Summary of Other Conclusions—What the Combined Evidence Says:  The pre-European fire regime at CMR was lightning driven, in contrast with some other parts of the country where Native Americans were the dominant force. This was true before and after the 1730 acquisition of horses by the Blackfeet. Prior to 1730 the Blackfeet and their precursors along the upper Missouri River likely used fire in the landscape in ways similar to those used by the Mandans in the early European contact period downstream but this would have only augmented slightly the background lightning ignition regime.  Ignitions are more frequent at CMR than in some other parts of the shortgrass prairie because of four local orographic ignition sources: the Little Rocky Mountains, the Judith Mountains, the Bearpaw Mountains and the Snowy Mountains. Ignitions from the four sources reach their maximum in the center of the refuge and this pattern has existed for thousands of years. Any burning by First Nations people was superimposed on this fixed lightning ignition pattern.  Fire frequency is lowest in the eastern third of CMR, where ignitions are infrequent and fire spread is limited by extensive fire filters and firebreaks created by poorly vegetated gumbo and badland soils.  Lightning ignition density overrides topography in control of fire frequency in the central portion of CMR.  Fires in the Sand Creek section are the result primarily of ignitions from the Judith and Snowy

82 Mountains and not cold fronts.  Global warming is known to be increasing on fire frequency (Westerling et al. 2006) but the killing of old-growth trees by fires at CMR is the result of increased fire intensity. The fire scar chronologies indicate that woody fuels increased greatly beginning with saturation of the landscape with sheep and cattle around 1900, and the replacement of bison with cattle and sheep has led to a doubling of the length of fire intervals, providing enough time for accumulation of lethal amounts of juniper, pine and fir saplings that would have been kept down under the natural fire frequency.

Speculative conclusions deserving further study  Beaver were abundant in some CMR creeks historically—the streams with better water supply— while the shorter coulees were too dry. It should be possible with more historical and on-the- ground work to map the stream sections that originally supported beaver wetlands.  Besides providing water and cover for ducks and other waterfowl, and providing inland sources of water for bison and the other original large ungulates, now supplied by stock ponds, beaver wetlands may have constituted critical foraging, shelter and watering habitat for young sharptail grouse during the months May to August. It is possible that this is the reason for decline in this species from their pre-European abundance.

“IS THIS NATURAL”

Arguments for natural versus anthropogenic causes of the severe fires of July 2006. YES: At Lost Creek, the fire scar record shows four previous severe fires, in 1702, 1834, 1882 and 1982 (severe in the fire scar chronology sense of percent of the recorded trees scarred by fire). It might be argued that since fire frequency is lower in shortgrass prairie than in tallgrass and other habitats with more annual rainfall, there is more time for accumulation of fuels and fires should be hotter. This should apply, however, only in places where there is the potential for growth of woody species like juniper since there is little additional increase in grass fuels beyond the first two years after fire. In the case of the pine breaks, there is a gradient of fire frequency with some areas, such as steep north-facing ravines, likely to be missed by one or several fires, so that by the time a fire does make its way there, the accumulated multi-storied woody vegetation formed by juniper and tree saplings below, and canopy above, fuels stand-destroying fires. This is almost certainly the case but should have occurred only in an isolated pocket here and there so that in the original landscape there was an all-age distribution of stands in such pockets. Within the areas burned by the July 2006 fires too few sheltered areas were spared to provide the pre-fire distribution of old trees.

NO:  The 1982-2006 fire interval of 24 years was the shortest for severe fires at the site. The others averaged 93 years apart. While only one case, this fits with the increase in frequency of severe fires such as has been seen across the West since around 1980.  The severe fires of the past scarred but did not kill the older trees (this is true at least for 1982 severe fire, there are no old standing or fallen trunks lying around as will be created by the 2006 fire). Some of the trees killed in 2006 dated to the 1600s and 1700s and survived the ‘severe” fire of the past three centuries.  1880-2007 photo pairs at Cow Creek show increase in tree cover on the prairie/river gorge margins and upper slopes.  1915-2001 repeat photo trios around the Judith Mountains east of Lewistown by Klement et al. (2001) show increasing woody cover through the twentieth century.  Demography of two study plots, one at Lost Creek on the south side of the Missouri river and one at CK Creek on the north side, both show only scattered, old ponderosa pines prior to European settlement. Establishment of an understory of dense new stems was initiated at the same time as the peak saturation of the landscape by sheep and cattle around 1900, with infill complete by

83 around 1940, after which the stands were too shady to permit further tree establishment. Increased stem density leads to much hotter fires.  Pre-burn photos show heavy juniper buildup under pine and Douglas fir on a site at Soda Creek that was 100% killed by the July 2006 fire.  The fire scar chronologies show more than 100% increase in the first half of the 20th century in length of fire interval over that of pre-European settlement.

The sum of nine different kinds of evidence at CMR leads to the conclusion that the fires of 2006 were well beyond the natural range of severity, with the principal cause being the reduction in fire frequency that led to infill of previously much more open stands of ponderosa pine by dense woody vegetation in the era 1900-1950. Since then, woody fuel has continued to increase. Modern efforts at fire suppression cannot be blamed in this situation since the problem was initiated over 100 years ago. The challenge, to prevent the loss of remaining old-growth stands containing virgin trees, will be finding ways to reintroduce fire carefully, under moist conditions, in order to clear out the understory fuel without killing the canopy trees. If removal of existing heavy fuels can be accomplished, at least in critical areas, future fires, natural or prescribed will be no threat to the trees.

BIBLIOGRAPHY AND LITERATURE CITED

Bailey, Robert G. 1997. Delineation of ecosystem regions. Environmental Mgmt 7:365-373.

Bahre, Conrad, Joseph. 1991. A legacy of change. University of Arizona Press, Tucson. 231 p.

Bent, Arthur Cleveland. 1963. Life histories of North American gallinaceous birds. Dover Press reprint, New York. p. 291-300.

Bureau of Agriculture (anon.) 1900. seventh report of the Bureau of Agriculture and Industry of the State of Montana. Helena, Montana: Independent Pub. Co., State Printers and Binders. Fergus p 174-5, Choteau, Valley

Cheney, Roberta C. 1983. Names on the face of Montana. Mountain Press, Missoula

DeVoto, Bernard. 1953. The journals of Lewis and Clark. Houghton Mifflin Co., Boston. 504 p.

Dorn, Robert D. 1984. Vascular plants of Montana. Mountain West Publishing, Cheyenne, WY. 276 p.

Dorn, Robert D. 2001. Vascular plants of Wyoming. Mountain West Publishing, Cheyenne, WY. 412 p.

Ewers, John C. 1958. The Blackfeet, raiders on the northwestern plains. University of Oklahoma Press, Norman, OK. 345 p.

Fergus, James. 1862-1902. Fergus Ledgers and Daybook, Stuart Collection, Montana Historical Society Library. Available in part at http://www.dangel.net/AMERICA/Fergus/James/CHAPTER%20XIII.html

Frost, Cecil C. 2006. Presettlement vegetation and natural fire regimes of Fort Bragg, North Carolina. Report for the U.S. Army, Fort Bragg, NC [with 2 GIS maps].

Frost, Cecil C. 2005. Presettlement vegetation and natural fire regimes of Fort Stewart, Georgia. Report for the U.S. Army, Hinesville, GA [with 2 GIS maps].

Frost, Cecil C. 2001 Presettlement vegetation and natural fire regimes of Camp Lejeune, Onslow County, North Carolina. Report for the U.S. Marine Corps, Camp Lejeune, NC. 170 p. [with 2 GIS maps]. Frost, Cecil C. 2000. Studies in landscape fire ecology and presettlement vegetation of the southeastern United States. Doctoral dissertation, University of North Carolina, Chapel Hill. 620 p.

Frost, Cecil C. 2000. Studies in landscape fire ecology and presettlement vegetation of the southeastern United States. Doctoral dissertation, University of North Carolina, Chapel Hill. 620 p.

Frost, Cecil C. 1998. Presettlement fire frequency regimes of the United States: a first approximation. Pages 70-81 in Theresa L. Pruden and Leonard A. Brennan (eds.). Fire in ecosystem management: shifting the paradigm from suppression to prescription. Tall Timbers Fire Ecology Conference Proceedings, No. 20. Tall Timbers Research Station, Tallahassee, FL.

Frost, Cecil C. 1993. Four centuries of changing landscape patterns in the longleaf pine ecosystem. Pages 17-43 in S.M. Hermann, ed. The longleaf pine ecosystem: ecology, restoration and management. Proc. Tall Timbers Fire Ecol. Conf. No. 18.

Frost, Cecil C., J. Walker and R.K. Peet. 1986. Fire-dependent savannas and prairies of the Southeast: original extent, preservation status and management problems. In: D.L. Kulhavy and R.N. Conner, eds. Wilderness and natural areas in the eastern United States: a management challenge. Nacogdoches, Texas: Center for Applied Studies, School of Forestry, Stephen F. Austin State University. p. 348-357.

Great Plains Flora Association. 1986. Flora of the Great Plains. University of Kansas Press, Lawrence, Kansas. 1402 p.

Guyette, Richard P. and Daniel C. Dey. 2000. Humans, Topography, and Wildland Fire: The ingredients for long-term patterns in ecosystems. p. 28-35 in: Daniel A. Yaussy, compiler, Proceedings: Workshop on Fire, People, and the Central Hardwoods Landscape. USDA Forest Service, Newtown Square, Pennsylvania.

Hammond, E.H. 1964. Classes of land surface form in the forty-eight States. Annals of the Association of American Geographers 54. Map supplement no. 4, 1:5,000,000.

David C. Hartnett, Karen R. Hickman and Laura E. Fischer Walter. 1966. Effects of bison grazing, fire, and topography on floristic diversity in tallgrass prairie. J Range Management 49:413-420

Hitchcock and Cronquist. 1973. Flora of the Pacific Northwest. University of Washington Press, Seattle, WA. 730 p.

Johnson, James R. and Gary E. Larson. 1999. Grassland plants of the northern Great Plains. South Dakota State Univ., College of Agr. and Biol. Sci., SD Agr. Expt. Sta., Brookings, SD. 288 p.

Kartesz, J.T. 1994. Synonymized checklist of the vascular flora of the United States, Canada and Greenland. Second edition. Timber Press, Portland, OR.

Kaul, Robert. 1986. Physical and floristic characteristics of the Great Plains. In: Great Plains Flora Association. Flora of the Great Plains. University of Kansas Press, Lawrence, Kansas. p. 7-10.

Kay, Charles E., and Randy T. Simmons. 2002. Wilderness and Political Ecology. Salt Lake City: University of Utah Press. 352 p.

85 Kay, Charles. 2000. Native Burning in Western North America: Implications for Hardwood Forest Management. p 19-27 in: Daniel A. Yaussy, compiler, Proceedings: Workshop on Fire, People, and the Central Hardwoods Landscape. USDA Forest Service, Newtown Square, Pennsylvania.

Kay, Charles E. and Dale L. Bartos. 2000. Ungulate herbivory on Utah aspen: assessment of long-term exclosures. J. Range Management 53:145-153

Klement, KD, RK Heitschmidt and CE Kay. 2001. Eighty years of vegetation changes in the northern Great Plains. USDA Agricultural Research Service Conservation Research Report No. 45 (14 historical repeat photo sets for the Judith Mountain region).

Larpenteur, Charles. 1872 [1989]. Forty years a fur trader on the Upper Missouri. The personal narrative of Charles Larpenteur 1833-1872. University of Nebraska Press Bison Book reproduction of the 1933 Lakeside Press classics edition by R.R. Donnelly & Sons, Chicago. 388p.

Larson, Gary E. and James R. Johnson. 1999. Plants of the Black Hills and Bear Lodge Mountains. South Dakota State Univ., College of Agr. and Biol. Sci., SD Agr. Expt. Sta., Brookings, SD. 608 p.

Lavin, Matt and Cathy Seibert. 2005. Grasses of Montana. Online pdf file from MSU (MONT) Herbarium, Bozeman, MT 65 p.

McHugh, Tom. 1972. The time of the buffalo. University of Nebraska Press, Lincoln, Nebraska. 339 p.

McGregor, Ronald L., and T.M. Barkley, eds. 1986. Flora of the Great Plains. University Press of Kansas, Lawrence, KS. 1402 p.

Mech, L. David. 1970. The wolf. University of Minnesota Press, Minneapolis. 384 p.

Murie, Olaus J. (1889-1963). The elk of North America .

Murie, Olaus J. 1935. Report on the Fort Peck Migratory Bird Refuge (now part of Charles M. Russell NWR). Unpublished report on a reconnaissance of the area to determine the advisability of establishing a refuge bordering the Fort Peck Dam project. 33 p.

Ordway, John. 1805 The journals of John Ordway. in: Vol. 9 of the journals of the Lewis and Clark expedition. Gary E. Moulton, ed. University of Nebraska Press, Lincoln, NE. 419 p.

Savage, Melissa and Thomas W. Swetnam. 1990. Early 19th-Century Fire Decline Following Sheep Pasturing in a Navajo Ponderosa Pine Forest. Ecology 71:2374-2378.

Sargent, C.S. 1884. Report on the forests of North America. USDI Census Office, Washington, DC. 612 p. (has some cattle, sheep, etc. from Censuses of Agriculture. From 1840 or 1850 on)

Schieman, Donald A. 2005 Wildflowers of Montana. Mountain Press, Missoula, MT 306 p.

Schultz, James Willard. 1902 [2003]. Floating on the Missouri. Riverbend Publishing, Helena, MT. 135 p.

Schultz, James Willard. 1906. My Life as an Indian. Houghton Mifflin Co.

Schultz, James Willard. 1962. Blackfeet and buffalo, memories of life among the Indians, University of Oklahoma Press, Norman, OK. 384 p. (A collection of tales, many previously published in magazines,

86 some never published and true stories of the past told to him by Indian friends, some with important historical information such as the acquisition of horses by the Blackfeet)

Stambaugh, Michael C. 2008. Predicting spatio-temporal variability in fire return intervals using a topographic roughness index. Forest Ecology and Management 254: 463-473.

Strom, Donald E. 1984. Soil Survey of McCone County. Washington, DC: USDA Soil Conservation Service.

Stubbendieck, James, Stephen L Hatch and LM Landholt. 2003. North American Wildland Plants (earlier editions were called N.A. Range Plants), 6th ed. Nebraska University Press, Lincoln, NE. 514 pages, 200 b/w illus, 21 figs. #140734W | ISBN: 0803293062

Vance, F.R., J.R. Jowsey, J.S. McLean and F.A. Switzer. 1999. Wildflowers of the northern Great Plains. University of Minnesota Press, Minneapolis, MN. 382 p.

Wambolt, Carl W. and Michael R. Frisina. 2002. Montana sagebrush guide. Montana Dept. of Fish, Wildlife & Parks, Bozeman, MT. 71 p.

Welch, Bruce L. 2005. Big sagebrush: a sea fragmented into lakes, ponds and puddles. USDA Forest Service Rocky Mountain Research Station Gen Tech. Rep. RMRS-GTR-144. 210 p.

Westerling, A.L., H. G. Hidalgo, D. R. Cayan, T. W. Swetnam. 2006. Warming and Earlier Spring Increase Western U.S. Forest Wildfire Activity. Science 313:940-943

Willmore, Illa. 2006. pers comm. local history. (prairie shacks and buffalo tracks) 464-5461

Willmore, Illa, ed. 1990. Homestead Shacks over Buffalo Tracks. Roy History-Geneology Society. Roy, Montana.

George W. Wingate. 1886 [1998]. Through the Yellowstone Park on horseback. Univ. of Idaho Press, Moscow, ID.

Wright, Henry A. and Arthur W. Bailey. 1982. Fire ecology: United States and Southern Canada. New York: John Wiley & Sons. 501 p.

Wright, Henry A., Leon F. Neuenschwander, Carlton M. Brewer. 1979. The role and use of fire in sagebrush-grass and pinyon-juniper plant communities, a state of the art review. USDA Forest Service, Intermountain Forest and Range Experiment Station Gen. Tech. Rep INT-58. 46 p.

87 APPENDIX 1. PRE-EUROPEAN VEGETATION TYPES OF THE CHARLES M RUSSELL NATIONAL WILDLIFE REFUGE

TABLE OF CONTENTS Page Using soils as a starting point for putting boundaries on vegetation ...... 1 Plant identification ...... 2 Vegetation type descriptions ...... 7 1.1 “Rock” Outcrops and Poorly Consolidated Geologic Sediments ...... 3 1.2 Rock Outcrop-Vegetation Mosaic ...... 3 1.3 Badlands ...... 4 1.4 Dwarf Artemisia tridentata Flats-Dry Ridgetop Prairie Mosaic ...... 5 2.1 Dry Needlegrass-Wheatgrass Prairie ...... 14 2.2 Mesa-top Meadows ...... 14 2.3 Dry Prairie-Juniper Shrubland Mosaic with Occasional Ponderosa Pine ...... 16 24 Bottomland flats, Playas, Saline Flats, Alkaline Big Sagebrush Flats ...... 16 2.5 Mesic and Dry-Mesic Bottomland Needlegrass-Wheatgrass Prairie with Occasional Cottonwoods ...... 17 2 6 Prairie Dog Colonies (not mapped) ...... 18 3.1 Ponderosa Pine Savanna ...... 20 3.2 Ponderosa Pine-Juniper-Dry Prairie Mosaic ...... 23 3.3 Ponderosa Pine Slopes with Prairie Meadows and Occasional Douglas Fir ...... 23 4.1 Colluvium and Toe Slope Shrublands ...... 24 4.2 High River and Stream Terrace Shrubland and Prairie ...... 24 5.1 Saline Dry Drains ...... 24 6.1 Intermittently Flooded Lakes, Marshes and Vernal Pools ...... 24 6.2 Wooded Wetland Ravines and Drains ...... 24 6.3 Small Stream Wetland Mosaic Structured by Beaver and Fire ...... 24 7.1 Low Terrace Cottonwood Flats ...... 24 7.2 River and Small Stream Wetlands ...... 24

LIST OF FIGURES Figure Page 1. Original vegetation and fuel types – western half of CMR ...... 2 2. Original vegetation and fuel types – eastern half of CMR ...... 3 3. Badlands seen from a hill on north side of Haxby Road ...... 5 4. Dwarf Artemisia fire “islands” ...... 12 5. Dry prairie dominated by needle-and-thread and big sagebrush...... 13 6 Dry shortgrass prairie with little bluestem ...... 15 7. UL Bend prairie grasslands with Mickey Butte in right background ...... 16 8. Lip of Mickey Butte mesa top, looking southwest toward U.L. Bend ...... 16 9. Juniperus horizontalis site on a north-facing slope burned in the Heartland Fire ...... 17 10 CK Creek mesic bottomland prairie ...... 21 11 Ponderosa pine parkland or savanna at Lost Creek ...... 22 12 Musselshell Trail patch mosaic of dry prairie, ponderosa pine ...... 23 13 Silver sage (Artemisia cana ) on colluvium ...... 25 14 Dry Lake, head of UL Bend, Phillips County ...... 26 15 Wooded drain on road between hwy 24 and McGuire Recreation Area ...... 26 16 Wet phase of cottonwood flats...... 27 17 High phase of cottonwood flats ...... 28

The following vegetation classification was used for making the GIS map. Under most of the vegetation groups, samples of one or more typical dominant vegetation types are listed, with the prefix CT for “community type”. The examples given are interpreted to be some of the natural types as they existed with bison and other native grazers under the natural fire regimes and before introduction of livestock or invasion of exotic species such as Japanese brome and sweet clover.

The special vegetation classification used for this study relies on physiography and field investigation of remaining natural vegetation and soils. The emphasis is on categorizing vegetation in relation to fuel type, fuel quantity and qualities of vegetation and landscape that relate to fuel continuity and fire spread. Some categories might have been given different names and split or lumped into other types in a more standard vegetation classification. That said, the resulting GIS layer can be used as a map of presettlement vegetation at CMR.

Each soil upon which each vegetation type occurs is listed with a two-letter prefix for the county in which they occur (FE190 = soil type 190 in Fergus county). There were 278 soil mapping units used within the boundaries of CMR. Some of these were simple series or consocies (consociations: soil mapping units composed of at least 85% of a single series) such as the Neldore clay, while most, such as the MC115 Neldore-Badland-Bascovy complex, included two or three units that soil mappers lumped because the individual polygons or patches of a particular soil type were too small to map individually at the scale they were using. The size of soil map units on the maps varied greatly among the six counties of CMR, with Fergus, Petroleum and Valley counties being the most coarse and with the newer McCone map being drawn at the finest scale.

I used these soil map polygons as a starting point to put boundaries on vegetation types, so the scale of vegetation units varies with the soil units. In many cases the vegetation types below represent a mosaic of vegetation on the different soils, so polygons in vegetation group 2.3 are labeled Dry Prairie-Juniper 2 Shrubland Mosaic with Occasional Ponderosa Pine (while those in group 3.2 are labeled Ponderosa Pine-Juniper-Dry Prairie Mosaic. Both are mosaics of essentially the same vegetation types but in the first prairie is dominant while in the second ponderosa pine is dominant. This refers to dominance on that mapped soil type as a whole in that particular county, but a few soil polygons may be found that have only a little pine. For mapping, the dominant type was determined from aerial photos and in many places with field observations or study plots on the ground.

Often there is not a simple one-to-one relationship between the soil series and vegetation. For a particular soil series it was sometimes necessary to put one polygon in one vegetation group and another of the same soil type into a different vegetation group. A few bottomland soil polygons were split and assigned to different vegetation types.

The goal was to create a general presettlement vegetation map of CMR with an eye toward vegetation as fuels. Using color infrared aerial photos it would be possible and desirable to have a more fine-scaled vegetation map. This would require hand-delimiting of types on aerial photos, followed by examination in the field and adjustment for any increases in woody cover resulting from historical reduction in fire frequency. This would be a substantial project in its own right and was beyond the scope of the fire frequency study. The descriptions below, however, are intended to represent the major vegetation types in the original landscape. Of these, only beaver ecosystem vegetation seems to have been unrecognizably altered from the original state.

Using soils as a starting point for mapping vegetation. Soils as delineated on the six NRCS soil maps were used to put boundaries on vegetation. That presented a number of problems. The mapping was done at different times and at widely different levels of detail. As mentioned above, of the six counties only McCone has been mapped at a level of resolution comparable to modern soil maps in more settled parts of the country. In the others large soil units were mapped where they could be described as a complex of three soil series. This occurred most often in the pine breaks and other areas where topography was complex. Often the same three series were presented as a new complex when a different member was dominant. These varying approaches resulted in the 278 soil mapping units on the refuge lands. In the classification below each map unit was assigned to one of 21 vegetation types. These were grouped according to their dominant vegetation types. These coarse map units contain up to three series, typically encompassing three or more vegetation types so such map units are described below as vegetation mosaics. Other vegetation map units may fit a single vegetation type.

Plant Identification. Taxonomy follows Dorn 2001, for most species, and Lavin and Seibert 2005 for grasses. The best method for identifying plants at CMR, using publications current as of 2007 seems to be to key plants using Dorn 1984, then check the species description in Flora of the Great Plains, then check for more modern nomenclature in Dorn 2001, Vascular plants of Montana. Grasses are best keyed using the online version of Grasses of Montana by Lavin and Seibert. The most useful picture books available as of 2006 seem to be Larson and Johnson’s Plants of the Black Hills and Bear Lodge Mountains, Vance et al.’s Wildflowers of the northern Great Plains, and Schieman’s wildflowers of Montana. There are at least a dozen other floras and picture books that are useful for a second opinion on plant descriptions and drawings or photos of plants .

Figure 1. Original vegetation and fuels – western half of CMR.

Figure 2. Original vegetation and fuels – eastern half of CMR.

The community types (CTs) listed under the vegetation groups below are just some examples of communities that I saw in my plots. There are others for each group.

GROUP 1. ROCK OUTCROPS, BADLANDS, SHALE AND GUMBO BARRENS 1.1 “Rock” Outcrops and Poorly Consolidated Geologic Sediments (mostly gray shale of the Bearpaw Formation in this category) ORIGINAL VEGETATION: Almost completely barren, rapidly eroding gray shale often with narrow bands of plants such as greasewood and saltbush along horizontal seams in strata where water is available or where soil texture permits root penetration. Grades into 1.2 below. CT: Sparse Sarcobatus vermiculatus-Artemisia tridentata-Atriplex spp. on steep, eroded slopes CT: Atriplex dioica in dense patches on VA4, nearly barren, soft gray clay from shale, in Valley County. The plants dry out and turn red by mid to late summer.

SOIL MAP CODES and SOIL SERIES FE190 “Rock” outcrop (gray clay from shale) PH1400F “Rock” outcrop (45%)-Arsite (35%) association, 8 to 60% slopes VA4 Badland VA55 “Rock” outcrop (gray clay from shale) 5

1.2 Rock Outcrop-Vegetation Mosaic (mostly on the Bearpaw shale in western half of CMR, some sandstones and siltstones in eastern half) ORIGINAL VEGETATION: Dry prairie grasses, forbs & low shrubs with barren patches caused by natural sheet erosion on slopes. Barrens interspersed with patches of ponderosa pine, sometimes with a few Douglas fir in fire-sheltered upland ravines and pockets places. CT: Artemisia tridentata-dry prairie graminoids and forbs CT: Pinus ponderosa- Juniperus scopulorum-Artemisia tridentata/dry prairie graminoids and forbs in naturally fire sheltered places CT: Pinus ponderosa-dry prairie graminoids and forbs SOIL MAP CODES and SOIL SERIES FE176 Neldore-Rock outcrop complex, 15 to 60 % PE66 Neldore clay-Rock outcrop complex, 15 to 45% slopes PH1972F Volborg (60%)-Rock outcrop (20%) association, 8 to 45% slopes

1.3 Badlands Distinguished from gray gumbo badlands above by light-colored, mixed-strata badlands. Paleocene Fort Union, late Cretaceous Hell Creek and Fox Hill Formation badlands as opposed to the older gray shales of the Bearpaw Formation. On the vegetation and fire frequency maps, soils in this group were used to define fire filter zones that reduced fire frequency in the original landscape. ORIGINAL VEGETATION: CT: Artemisia tridentata-dry prairie graminoids and forbs CT: Juniperus scopulorum-Artemisia tridentata-dry prairie graminoids and forbs CT: Sarcobatus vermiculatus-Atriplex spp-dry prairie graminoids and forbs

SOIL MAP CODES and SOIL SERIES MC7 Badland GA3F Badland GA1007F Badland (70%)-Cambeth association, 15 to 70 percent slopes (Hell Cr FM) MC7 Badland MC32 Cabbart-Badland complex, 15 to 45 % slopes MC94 Hillon-Badland, 15-45% MC165 Yawdim-Badland-Cabbart association MC166 Yawdim-Badland-Gerdrum association MC115 Neldore-Badland-Bascovy complex, 15 to 45 % slopes (JUSC-JUHO-Gumbo) GA3F Badland GA203F Neldore-Rock outcrop, soft-Bascovy complex, 15 to 45 percent slopes GA362F Cabbart-Rock outcrop, soft-Delpoint complex, 15 to 50 percent slopes GA369F Cabbart-Rock outcrop, soft (30%)-Yawdim complex, 15 to 70 percent slopes GA992F Yawdim-Rock outcrop, soft-Cabbart association.. 15 to 45 percent slopes GA993F Yawdim-Badland-Gerdrum association, 15 to 45 percent slopes GA1002F Yamacall-Rock outcrop, soft (30%-Kobase association, 8 to 70 percent slopes GA1007F Badland (70%)-Cambeth association, 15 to 70 percent slopes (Hell Cr FM) GA1020F Cambeth (40%)-Rock outcrop, soft (20%)-Yawdim (15%)association, 25 to 70% slopes

Figure 3. Badlands seen from a hill on north side of Haxby Road, Garfield County. Note the abundance of Artemisia tridentata on the side slopes in this natural fire refugium, silver sage in the ravine bottom and the scarcity of any grass to carry fire except on the small flats in the middle distance. The thin blue line on the horizon is the clay ridge separating the lake from the Willow Creek flats to the north.

1.4 Dwarf Artemisia tridentata Flats-Dry Ridgetop Prairie Mosaic ORIGINAL VEGETATION: The study plot seen from above in Figure 4 below was dominated almost entirely by “big” sagebrush. The plants were of all sizes with a density of 27,100 stems per hectare, the greatest I have seen anywhere. The unusual density results in part from the small size of many of the plants and from the lack of competition. Of 15 other plant species present only junegrass (Koeleria micrantha and Bouteloua hirsuta exceeded 5% cover. CT: Artemisa tridentata- very sparse dry prairie grasses and forbs SOIL MAP CODES and SOIL SERIES PE38 Gerdrum-Bascovy clays, 2-15% slopes – CMR17 dwarf ARTR untouched by fire PE40 Gerdrum clay-Vanda silty clay, 1-6% slopes (Gerdrum in PE is high pH, saline clay)

Figure 4. Dwarf Artemisia fire “islands” along road 315 on uplands above Soda Creek are the unburned light gray areas in a mostly burned landscape. These sterile upland flats support dense but tiny Artemisia tridentata but too little grass to carry fire. This is a natural phenomenon and not an artifact of grazing. Taken a few days after the July 15, 2006 lightning fire complex.

GROUP2. PRAIRIES While some of the more lush areas, including those with considerable little bluestem, such as occurs in some of the McCone County portion of CMR and some of the area between Mickey Butte and UL Bend, might be considered mixed-grass prairie, most authors designate the region around CMR as shortgrass prairie: Kaul, Robert. 1986. Physical and floristic characteristics of the Great Plains. In: Great Plains Flora Association. Flora of the Great Plains. University of Kansas Press, Lawrence, Kansas. p. 7-10. Kaul showed CMR consisting of short grass prairie: grama, needle and wheat grasses (Bouteloua-Stipa- Agropyron), with evergreen coniferous forest along the river (Kaul 1986, p.8). Bailey, Robert G. 1997. Delineation of ecosystem regions. Environmental Mgmt 7:365-373, placed it in his short-grass category 3111: 3000 Dry Domain 3100 Steppe Division 3110 Great Plains Short-grass Prairie Province 3111 Grama-Needlegrass-Wheatgrass Section 3112 Wheatgrass-Needlegrass Section 3113 Grama-Buffalograss Section

2.1 Dry Needlegrass-Wheatgrass Prairie ORIGINAL VEGETATION: CT: Pascopyron smithii-Pseudoregneria spicatum-diverse graminoids and forbs CT: Nasella viridulus-mixed dry prairie graminoids and forbs 8 CT: Hesperostipa comata-Aremisia tridentata (on infrequently burned, dry, fine sandy loams) CT: Schizachyrium scoparium- Nasella viridulus -mixed wheatgrass spp-mixed forbs of prairies with sandy loam soils (especially in McCone and southern Garfield Counties)

Figure 5. Dry prairie strongly dominated by needle-and-thread (Hesperostipa comata), with scattered big sagebrush on Busby-Twilight fine sandy loams, 2 to 8 percent slopes . Sandy loam soils are typically avoided by prairie dogs, and this plot was adjacent to a large prairie dog town on a denser soil (Floweree-Cambeth silt loams, 2 to 8 percent slopes) just behind the person in the distance. Access to fire is reduced by badlands soils nearby. The dominance of needle-and-thread may be the result of the soil type along with low incidence of fire in this part of the refuge because reduction in fire frequency often leads to dominance by one or two species. Nelson Creek Recreation Area, on the south side of Nelson Creek near its mouth on Big Dry Creek.

Figure 6. Dry shortgrass prairie with little bluestem (Schizachyrium scoparium) as one of the patch dominants (easily distinguished by its reddish color in August). Just north of the McGuire Creek Recreation Area, McCone County.

Figure 7. UL Bend prairie grasslands with Mickey Butte in right background, Brandon Butte in distance on left. The prairie soil type in foreground is PH1251E, Neldore- Bascovy clays, 8-35%, range grass productivity about 1000 lbs/acre. The grass density at this long-ungrazed spot is clearly enough to support rapid fire spread. The zone of juniper in the middle distance has been missed by several fires in recent decades.

SOIL MAP CODES and SOIL SERIES FE113 Gerdrum (clay loam)-Absher (clay loam)complex, 2 to 8 % slopes FE174 Neldore-Thebo clays, 4 to 25 % slopes FE233 Thebo clay 0-8% slopes FE234 Thebo clay 8-25% slopes FE235 Thebo-Weingart-Absher clays, 4 to 15 % slopes FE175 Neldore-Thebo clays, 25-60% slopes PE2 Abor-Neldore silty clays, 2-8% slopes (SM18) PE8 Bascovy-Neldore silty clays, 2-15% slopes PE18 Cabbart-Delpoint-Rock outcrop 8-45% slopes (both from sandstone substrate in VA) PE20 Cabbart-Yawdim-Delpoint 15-35% slopes (N tip, Chain Buttes) PE32 Delpoint-Cabbart-Yamac loams, 4-15% slopes SM21 PE34 Ethridge clay loam, 2-8% slopes PE36 Evanston loam, 2-8% slopes (SM25 1 polygon) PE39 Gerdrum-Creed, 1-6% slopes PE41 Gerdrum-Vanda-Creed, 1-8% slopes (SM18) PE61 Neldore silty clays, 4-25% slopes (SM5 on small mesa) PE62 Neldore-Abor silty clays, 4-15% slopes (SM5 on small mesa) PE65 Neldore-Neldore saline silty clays, 4-25% slopes PE98 Yamac-Delpoint loams, 2-8% slopes (SM21) PH96D Megonot (50%)-Kobase (35%) silty clay loams, 8 to 15% slopes PH97D Neldore (60%)-Bascovy (30%) clays, 4 to 15% slopes PH250E Bascovy (40%)-Neldore (30%)-Weingart (20%) clays, 8 to 25% slopes PH251C Bascovy (55%)-Neldore (30%) clays, 2 to 8% slopes PH301C Marvan complex (Marvan saline 50%, Marvan 30%), 2 to 8% slopes 11 PH521B Elloam (50%)-Absher (40%) complex, 0 to 4% slopes PH791E Yamacall complex (Yamacall calcareous 60%, Yamacall 30%), 8 to 35% slopes PH924E Sunburst (35%)-Bascovy (30%)-Neldore (20%) complex, 8 to 35% slopes PH925C Sunburst (35%)-Bascovy (30%)-Weingart (20%) complex, 2 to 8% slopes PH1052B Elloam (50%)-Thoeny (30%) association, 0 to 6% slopes PH1251E Neldore (50%)-Bascovy (25%) association, 8 to 35% slopes PH1332C Phillips (35%)-Elloam (25%)-Thoeny (20%) association, 0 to 8% slopes PH1920F Sunburst (40%)-Neldore (30%) association, 15 to 45% slopes PH1972F Volborg (60%)-Rock outcrop (20%) association, 8 to 45% slopes VA8 Elloam gravelly clay, 2-9% slopes concave, SM277 VA11 Evanston loam, 2-9% slopes VA12 Evanston loam, sandstone substrate, 2-5% slopes. soft sandstone at 40-60” VA15 Evanston-Marmarth loams, 3-12% slopes (both from sandstone substrate) VA39 Marmarth-Cabbart loams, 5-25% slopes (both from sandstone substrate) VA47 Phillips loam-Elloam clay, 1-9% slopes VA57 Scobey clay loam, 1-9% slopes, formed in glacial till VA59 Scobey-Sunburst clay loams, 5-25% slopes both formed in glacial till VA60 Sunburst clay loam, 9-35% slopes, formed in glacial till VA61 Sunburst clay loam-Lisam clay, 9-35% slopes, formed in gray shale wi a thin layer of glacial till VA68 Thebo-Lisam clays, 2-15% slopes (with 10% patches of gray gumbo) Thebo formed in shale, Lisam formed in clay & thin glacial till MC5 Alona silt loam, 0 to 8 % slopes from alluvium MC9 Barkof silty clay, 2-8% MC10 Bascovy silty clay, 2 to 8 % slopes from shale MC18 Busby fine sandy loam, 2 to 8 % slopes MC19 Busby fine sandy loam, 8 to 15 % slopes MC20 Busby-Fleak complex, 15 to 45 % slopes (SCSC) MC21 Busby-Twilight fine sandy loams, 2 to 8 % MC22 Busby-Twilight-Fleak complex, 8 to 15% slopes MC23 Busby-Yamac-Fleak complex, 15 to 45% slopes (SCSC) MC24 Busby-Yetull fine sandy loams, 2 to 8% slopes MC31 Cabbart silt loam, 15 to 25 % slopes MC33 Cabbart-Kirby complex, 8 to 45 % slopes MC34 Cabbart-Twilight complex, 15 to 45% slopes MC35 Cabbart-Yawdim 4-15% slopes MC36 Cabbart-Yawdim complex, 15 to 45% slopes MC41 Cambeth silt loam, 2 to 8 % slopes MC42 Cambeth-Cabbart silt loams, 8 to 15 % slopes MC43 Cambeth-Twilight-Cabbart complex, 4 to 15% slopes MC60 Evanston loam, 0 to 2 % slopes MC61 Evanston loam, 2 to 8 % slopes MC46 Chinook fine sandy loam, 0 to 4% slopes MC47 Chinook fine sandy loam, 4 to 8% slopes MC48 Chinook fine sandy loam, 8 to 15% slopes MC49 Chinook fine sandy loam, gullied, 2 to 8% slopes MC50 Creed loam, 0 to 8% slopes MC51 Creed-Gerdrum complex MC58 Ethridge silty clay loam, 0 to 4% slopes SM62 MC59 Ethridge silty clay loam, 4 to 8% slopes MC62 Evanston-Gerdrum complex, 2 to 8% slopes MC63 Farland silt loam , 0-4% slopes (couldn’t find) MC65 Floweree silt loam, 0-4% slopes MC66 Floweree silt loam, 4 to 8% slopes MC67 Floweree-Cambeth silt loams, 2 to 8% slopes MC68 Gerdrum clay loam 12 MC70 Gerdrum-Absher clay loams, 0 to 8% slopes MC71 Gerdrum-Yawdim-Fleak complex, 0 to 8% slopes MC91 Hillon loam, 2 to 8% slopes MC92 Hillon loam, 8 to 15% slopes MC95 Hillon-Yamac-Fleak complex, 15 to 45 % slopes MC97 Kremlin loam, 0 to 4 % slopes MC98 Kremlin loam, 4-8% slopes MC105 Lonna silty clay loam, 0-4% (Big Dry Creek) MC108 Macar loam, 4-8% MC120 Rominell loam, 0 to 8 % slopes MC121 Rominell loam, gullied, 0 to 8 % slopes MC128 Sunburst clay loam, 2 to 8 % slopes MC129 Sunburst clay loam, 8 to 15 % slopes MC130 Sunburst clay loam, 15 to 45 % slopes MC133 Telstad loam, 2 to 8 % slopes SM45 MC134 Telstad-Hillon loams, 2 to 8 % slopes MC135 Telstad-Hillon loams, 8 to 15 % slopes MC142 Twilight-Yetull fine sandy loams, 8 to 15% slopes MC148 Vanda clay, 0 to 8 % slopes MC153 Wabek sandy loam, 4 to 15 % slopes MC154 Wabek sandy loam, 15 to 45 % slopes MC155 Weingart clay, 2 to 8 % slopes MC159 Yamac loam, 0 to 4 % slopes MC160 Yamac loam, 4 to 8 % slopes MC161 Yamac loam, 8 to 15 % slopes MC162 Yamac-Twilight complex, 2 to 8 % slopes MC163 Yamac-Twilight-Fleak complex, 8 to 15 % slopes GA12B Antwerp silty clay loam, 0 to 4 percent slopes GA20E Neldore-Abor silty clays, 15 to 35 percent slopes York Island GA31C Busby fine sandy loam, 2 to 8 percent slopes GA31D Busby fine sandy loam, 8 to 15 percent slopes GA36E Cabbart silt loam, 15-25% GA38C Chinook fine sandy loam, 2 to 8 percent slopes GA53C Foreleft-Gerdrum complex, 2 to 8 percent slopes GA63C Hillon loam, 2 to 8 percent slopes GA66D Kobase silty clay loam, 8 to 15 percent slopes GA72A McKenzie silty clay, 0 to 2 percent slopes, 1 small polygon, PD now (a lake flooding artifact) GA74A Marvan silty clay, 0 to 2 percent slopes GA74C Marvan silty clay, 2 to 8 percent slopes GA86C Archin loam, 2 to 8 percent slopes GA90C Sonnett silty clay loam, thin surface, 2 to 8 percent slopes GA93C Telstad loam, 2 to 8 percent slopes GA95C Weingart clay, 2 to 8 percent slopes GA96B Vanda silty clay, 0 to 4 percent slopes GA97D Vendome sandy loam, 4-15% (Aridic Haplustolls) GA98C Yamacall loam, 4-8 percent slopes (remnant upland flat) GA205D Neldore-Bascovy complex, 2 to 15 percent slopes (1 soil polygon) GA313D Busby-Twilight fine sandy loams, 2 to 15 percent slopes GA313E Busby-Twilight-Blacksheep fine sandy loams, 8 to 35 percent slopes GA314F Busby-Yamacall-Fleak compiex, 15 to 45 percent slopes GA316F Twilight-Blacksheep fine sandy loams, 15 to 70 percent slopes GA318D Busby-Twilight-Fleak complex, 8 to 15 percent slopes GA352F Cabbart- Twilight complex, 15 to 45 percent slopes GA374E Cambeth-Cabbart-Yawdim complex, 15 to 25 percent slopes GA376C Cambeth-Cabbart silt loams, 2 to 8 percent slopes 13 GA376D Cambeth-Cabbart silt loams, 4 to 15 percent slopes GA378C Cambeth silt loam, noncalcareous, 2 to 8 percent slopes GA381C Chinook-Kremlin complex, 2 to 6 percent slopes GA385D Chinook fine sandy loam, 8 to 15 percent slopes GA391C Creed-Gerdrum complex, 0 to 8 percent slopes GA435E Delpoint-Cabbart-Yawdim complex, 4 to 25 percent slopes GA521C Floweree-Cambeth, silt loams, 2-8% SM46 Big Dry Cr. GA541C Gerdrum-Yawdim-Fleak complex, 0 to 8 percent slopes GA552B Gerdrum-Creed complex, 0 to 4 percent slopes GA553B Gerdrum-Vanda silty clays, 0 to 4 percent slopes GA555C Gerdrum-Marvan silty clays, 2 to 8 percent slopes GA704D Lonna, Cambeth, and Yamacall soils, 8 to 15 percent slopes, gullied (Big Dry Creek) GA792D GA Neldore-Abor silty clays, 2 to 15 percent slopes York Island GA796D Abor-Neldore silty clays, 4 to 15 percent slopes GA915C Busby-Yetull fine sandy loams, 2 to 8 percent slopes GA941E Cabbart-Havre loams, 0 to 35 percent slopes GA862C Rominell loam, 0-8%, gullied GA864C Rominell loam, 0 to 8 percent slopes GA916D Twilight-Yetull fine sandy loams, 8 to 15 percent slopes GA931D Telstad (Aridic Argiustolls)-Hillon loams, 8 to 15 percent slopes GA982D Yamacall-Delpoint-Cabbart loams, 8 to 15 percent slopes GA983D Yamacall-Twillght-Blacksheep complex, 8 to 15 percent slopes GA985C Yamacall-Busby complex, 2 to 8 percent slopes GA986C Yamacall-Twilight complex, 2 to 8 percent slopes GA987D Yamacall-Twllight-Fleak complex, 8 to 15 percent slopes

2.2 Mesa-top Meadows ORIGINAL VEGETATION: CT: Nassella viridula-Krascheninnikovia lanata In these areas, long ungrazed by cattle, there was dense cover of green needlegrass and other species. Native ungulates observed while doing study plots were primarily bighorn sheep and mule deer. On Mickey Butte there were a high number of stems of winterfat, the most seen in any plot but the plants were small, typically around 10-20 cm tall. The pattern seen in study plots suggest that winterfat is grazed but not killed by natives while it is eliminated in areas grazed by cattle (exact mechanism unknown). SOIL MAP CODES and SOIL SERIES PH1373C Evanston (40%)-Chinook (25%)-Marmarth (20%) association, 0 to 8% slopes PH1523C Elloam (40%)-Phillips (25%)-Absher (15%) association, 0 to 8% slopes (UL Bend PDT) GA701C Lonna-Cambeth silt loams, 2 to 8 percent slopes

Figure 8. Lip of Mickey Butte mesa top, looking southwest toward U.L. Bend. Dominant grass on much of the top is green needlegrass (Nasella viridula). Note close-cropped grass: this was typical of a zone about 10 meters wide near the mesa edge. Dense rodent droppings about 1-2 cm long were found in this zone (pack rats?). The rest of the surface toward the interior had dense grass cover about 20-30 cm tall. Technically, any flat-topped erosion remnant of flat-lying sedimentary deposits with a top >1000 meters square is a mesa. In contrast a butte tends to be rounded, with any flat top being smaller than 1000 square meters, and a butte may be of sedimentary, metamorphic or igneous origin as with the properly named Black Butte. To complicate matters, the term mesa (Spanish origin) is applied more widely in the Southwest while butte (French origin) is used more broadly in the Northwest.

2.3 Dry Prairie-Juniper Shrubland Mosaic with Occasional Ponderosa Pine (within its range which ends at Fourth Ridge). East of Fourth Ridge in Valley county and in all of McCone county there is no ponderosa so this mosaic is composed of just prairie grasses and juniper. ORIGINAL VEGETATION – Many community types possible, using the following Patch Components: Oryzopsis hymenoides (on narrow, dry ridgetops, especially where sandy) Pascopyron smithii-Pseudoregneria spicatum-diverse graminoids and forbs Nasella viridulus-mixed dry prairie graminoids and forbs Hesperostipa comata-Aremisia tridentata Schizachyrium scoparium- Nasella viridulus -mixed wheatgrass spp-mixed forbs of prairies with sandy loam soils Juniperus scopulorum 15 Juniperus horizontalis Pinus ponderosa

Figure 9. Juniperus horizontalis site on a north-facing slope burned in the Heartland Fire, 2005. The dark, round circles (with small remnants of juniper branches) are the fire shadows of burned juniper. There were one or two green branches of creeping juniper, apparently from post-fire sprouts. On private land just north of the refuge boundary at Iron Stake Ridge (seen in left background). CT before the fire: Juniperus horizontalis-Juniperus scopulorum/Nasella viridulus-Koelaria macrantha/Selaginella densa-diverse dry prairie grasses and forbs. 67 species were seen in a 1/10 hectare plot, unusually high species diversity for shortgrass prairie.

SOIL MAP CODES and SOIL SERIES PH1021E Cabbart (40%)-Twi1ight (30%)-Yawdim (15%) association, 8 to 35% slopes (dry prairie-juniper- PIPO woodland mosaic, 1 cluster region near Iron Stake Ridge, sparse patches of PIPO, dry prairie wi JUHO) PH1920F Sunburst (40%)-Neldore (30%) association, 15 to 45% slopes VA6 Cabbart-Delpoint loams, 9-35% slopes. Ponderosa pine beginning at Fourth Ridge and west MC11 Bascovy-Sunburst complex, 15 to 45 % slopes from shale MC116 Neldore-Bascovy complex, 2 to 15 % slopes MC117 Neldore-Yamac-Badland complex, 15 to 45 % slopes (peak W of 24 wi L&C Milk River overlook) GA204F Neldore-Yamacall-Rock outcrop, soft (25%), complex, 15 to 45 percent slopes GA312F Busby-Fleak complex, 15 to 45 percent slopes, (Juniper enhanced from natural levels by fire protection from lake) GA792E Neldore-Abor silty clays, 8 to 15 percent slopes Haxby Point

2.4 Bottomland flats, Playas, Saline Flats, Alkaline Big Sagebrush Flats ORIGINAL VEGETATION: CT: Atriplex spp-mixed graminoids of alkaline and saline flats. CT: Artemisia tridentata SOIL MAP CODES and SOIL SERIES PH41A Vaeda clay, 0 to 2% slopes PH302B Marvan (Marvan saline 50%)-Vanda (35%) clays, 0 to 8% slopes. Individual mapped soil polygons mapped as 302B in Phillips county were assigned to three different vegetation types). PH402B Gerdrum (50%)-Absher (40%) complex, 0 to 4% slopes PH604A Bullhook loam 0-2% fine-loamy, mixed (calcareous) frigid Ustifluvents (alkaline, saline) SM111 MC12 Benz clay loam, 0 to 8 % slopes

2.5 Mesic and Dry-Mesic Bottomland Needlegrass-Wheatgrass Prairie with Occasional Cottonwoods ORIGINAL VEGETATION: This type occurs in small stream bottomlands on soils that are in the bottom but are too high to have been flooded by beaver impoundments. Because of the more loamy bottomland soil texture and higher moisture availability, grass productivity is among the highest of soils at CMR.

Figure 10. CK Creek mesic bottomland prairie. Vegetation consists of a patch mosaic of grasses and big sagebrush. Dense patches of big sagebrush would have been natural in some of the fire sheltered loops of the stream. The lower, greener bottom meandering along the cut banks of the creek was may have been filled with beaver wetlands historically. The eroding cut bank along the creek seen in left center may be a result of the removal of beaver.

17 CT: Artemisia tridentata/mixed wheatgrass species-mesic bottomland forbs. CT: Agropyron (Pascopyrum) smithii-mixed wheatgrass species-mesic bottomland forbs. CT: Populus deltoides/diverse bottomland grasses and forbs

SOIL MAP CODES and SOIL SERIES PE46 Harlem-Havre, 0-2% slopes, occ flooded PE52 Havre-Harlem, 0-2% slopes, occ flooded, channeled PE86 Vanda-Harlem-Marvan silty clay, 0-4% slopes PE 60 (only the one polygon where it occurs in Dry Coulee, not along the Missouri River) PH903A Harlake (50%)-Lostriver (40%) clays, 0 to 2% slopes (lower Hawley Creek) PH1090B Harlake (45%)-Marvan (30%) association, 0 to 4% slopes (Jim Wells Creek, Siparyann) MC56 Dimmick silty clay MC69 Gerdrum clay loam, gullied, 8 to 15 % slopes MC72 Gerdrum-Yawdim-Fleak complex, 8 to 45 % slopes MC79 Hanly loamy fine sand MC180 Vendome sandy loam, 4-15% (Haplustolls)(loc?) GA60A Harlake silty clay, 0 to 2 percent slopes, rarely flooded (Squaw Creek, Hay Coulee) GA613A Havre and Glendive soils, frequently flooded, channeled (Devil’s Creek) GA942A Havre-Bigsandy loams, 0 to 2 percent slopes, frequently flooded (Germaine Coulee) VA75 Ustic Torrifluvents, gently sloping (recent alluvium on stream terraces, loam & clay, WD to MWD but freq flooded, some are strongly alkaline, but not saline?)

2.6 Prairie Dog Colonies (not mapped) ORIGINAL VEGETATION: Prairie dog towns at CMR occur on a certain set of preferred soils (mostly textures finer than fine sandy loams) and a set of preferred slope and landscape positions (mostly slightly concave sites with slope <15%). Vegetation, soils and slope and landscape position data were collected for most of the colonies at CMR and will be summarized elsewhere. Both vegetation species composition and dominants vary highly between sites as illustrated by results from a sample of the 1/10 hectare vegetation plots below. Grass cover varies from none at all to 50%. CT: Plantago elongata (50-75% cover). This plot had only 10 plant species and no grass (plot CMR07-09, 7/12/07). CT: Plantago patagonica (50-75% cover)-Artemesia frigida (10-25% cover). 16 plant species and <5% grass cover (plot FE03, 8/18/07). CT: Picradeniopsis oppositifolia (75-95% cover) – Lepidium viginianum (10-25% cover) – diverse dwarf forbs and sparse shortgrass spp. of prairie dog towns (24 species, plot CMR07-05 7/6/07) CT: Picradeniopsis oppositifolia (50-75% cover) – mixed dwarf forbs of prairie dog towns (24 species, plot CMR07-08 7/11/07). No grasses but one small clump of Calamovilfa. At Jim Wells cabin, U.L. Bend CT: Poa secunda (25-50% cover) - Bouteloua hirsuta (25-50% cover)-diverse graminoids and dwarf forbs of prairie dog towns (35 species, plot CMR07-06 7/7/07) CT: Pascopyron smithii (25-50% cover) – Salsola iberica (25-50% cover)-diverse graminoids and dwarf forbs of prairie dog towns (26 species, plot PDT21 7/28/07) CT: Sphaeralcea coccinea (25-50% cover)- diverse graminoids and dwarf forbs of prairie dog towns (36 species, plot CMR07-03 6/27/07) CT: Hedeoma hispidum (25-50% cover)- mixed graminoids and dwarf forbs of prairie dog towns

GROUP 3. PONDEROSA PINE 3.1 Ponderosa Pine Savanna

18 ORIGINAL VEGETATION: This kind of open community of ponderosa pine over grass has been called parkland, woodland or savanna. The natural fire frequency is too high for Douglas fir or juniper. The few shrubs present are those that are able to resprout quickly after fire such as silver sage (Artemisia cana) and shunkbush (Rhus trilobata) and even these tend to be limited to slightly fire sheltered spots in the landscape. Grass cover here on sandy loam soils is lush and continuous promoting rapid and thorough spread of fire.

Figure 11. Ponderosa pine parkland or savanna (scattered trees over grass) at Lost Creek, in an area with the highest fire frequency area at CMR.

CT: Pinus ponderosa/Schizachyrium scoparium-Aristida purpurea-Nassella (Stipa) viridulus –Koelaria macrantha-diverse pine savanna-woodland grasses and forbs CT: Pinus ponderosa/Nassella (Stipa) viridulus –Koelaria macrantha-diverse pine savanna-woodland grasses and forbs CT: Pinus ponderosa/Echinacea angustifolia-diverse pine savanna-woodland grasses and forbs

SOIL MAP CODES and SOIL SERIES: PH221D Hillon gravelly loam (50%)-Kevin gravelly lay loam (35%) complex, 8 to 15% slopes, gravelly loam & gravelly clay loam from glacial till PH221E Hillon (55%)-Kevin (30%) complex, 15 to 25% slopes GA311D Busby-Blacksheep- Twilight fine sandy loams, 2 to 15 percent slopes (Lost Creek) GA373C Cambeth, noncalcareous-Megonot complex, 2 to 8 percent slopes (Devils Creek) GA375D Cambeth- Twilight-Cabbart complex, 4 to 15 percent slopes (Lost Creek, Big Dry Cr.) GA382D Chinook-Twilight fine sandy loams, 2 to 15 percent slopes GA383C Chinook- Twilight-Eapa complex, 2 to 8 percent slopes GA702D Lonna-Cambeth-Cabbart silt loams, 4 to 12 percent slopes 19 GA841C Ralph-Brushton silt loams, 2 to 8 percent slopes (Devil’s Creek) GA981C Yamacall-Delpoint loams, 2 to 8 percent slopes

3.2 Ponderosa Pine-Juniper-Dry Prairie Mosaic with dominants depending on location in the landscape. ORIGINAL VEGETATION:

Figure 12. View from Musselshell Trail (refuge road 315) on ridge leading down to the mouth of Soda Creek. Patch mosaic of dry prairie, ponderosa pine in slightly sheltered sites in foreground and on north-facing slopes on the south side of Soda Creek in the distance, and with juniper sheltered under the ponderosas and on the side slopes in right midground. This scene has perhaps 50% more woody cover than it would have had before 1890 because of reduced fire frequency (so this site was mapped in vegetation group 2.1. Most of the juniper on the right (west) side of photo was killed in the July 15, 2006 fire at this site.

20 SOIL MAP CODES and SOIL SERIES PH1970F Neldore (35%)-Bascovy (30%)-Rock (rock + gumbo) outcrop (20%) association, 8 to 60% slopes PH1971F Yawdim (40%)-Cabbart (30%)-Rock outcrop (15%), 25 to 70% slopes (Mickey Butte pediment) VA34 Lisam-Dilts clays, 3-35% slopes (all prairie east of Fourth Ridge) (no ponderosa at all in MC) GA25F Neldore-Yawdim silty clays, 15 to 60 percent slopes GA431E Delpoint-Yamacall-Cabbart loams, 8 to 25 percent slopes GA432F Delpoint-Cabbart-Yawdim complex, 25 to 70 percent slopes GA434D Delpoint-Busby-Blacksheep complex, 4 to 15 percent slopes GA371F Cambeth-Cabbart-Rock outcrop, soft, complex, 8 to 45 percent slopes GA991F Yawdim-Cabbart-Kobase complex, 15 to 70 percent slopes GA795E Weingart-Neldore complex, 4 to 25 percent slopes GA1018F Neldore-Cabbart-Blacksheep association, 15 to 60 percent slopes GA1022D Lonna-Cabbart-Cambeth association, 4 to 15 percent slopes

3.3 Ponderosa Pine Slopes with Prairie Meadows and Occasional Douglas Fir Ravines. ORIGINAL VEGETATION: Similar to 3.2 but dominated by Ponderosa pine. Fire influenced, but with reduced fire effects and frequency on more fire-sheltered lower slopes, flats and steep-sided ravines. (see photo pairs on pages 10 and 11 in text section). SOIL MAP CODES and SOIL SERIES FE64 Dilts-Julin-Rock outcrop complex, 15 to 50% slopes FE65Dilts-Thebo-Neldore clays, 4 to 60 % slopes PE9 Bascovy-Neldore-moist Neldore silty clays, 6-60% slopes PE16 Cabbart-Delpoint loams, 4-15% slopes PE23 Cabbart-moist Blackhall-Delpoint 6-60% slopes PE64 Neldore-Bascovy-Rock outcrop, 6-60% slopes PE70 Neldore-moist Bascovy-Neldore south silty clays, 6-60% slopes PE72 Pinebreaks-moist Neldore silty clays, 15-60% slopes PH973E Neldore, coo1 (55%)-Bascovy (30%)clays, 8 to 35% slopes (PIPO) PH974F Pinebreaks (50%)-Neldore (35%) clays, 15 to 60% slopes PH1021E Cabbart (40%)-Twi1ight (30%)-Yawdim (15%) association, 8 to 35% slopes (dry prairie-PIPO woodland mosaic, 1 cluster, sparse patches of PIPO, dry prairie wi JUHO) PH1066D Twilight (35%)-Cabbart (25%)-Marmarth (15%) association, 4 to 15% slopes PH1850F Cabbart (30%)-Twilight (30%)-Delpoint (25%) association, 25 to 70% slopes PH1973F Neldore, cool (35%) (-Neldore (30%)-Rock outcrop (20%), 15 to 60% slopes PH1976F Neldore (35%)-Pinebreaks (25%)-Bascovy (20%) association, 15 to 60% slopes PH1977F Volborg (40%)-Pinebreaks (25%)-Rock outcrop (15%) association, 15 to 60% slopes PH2972F Volborg (35%)-Neldore (30%)-Rock outcrop (15%) association, 15 to 60% slopes (the only ponderosa-dominant stands in VA are included in 3.2 mosaic above) (no ponderosa at all in MC) GA24F Neldore-Volborg, wooded, silty clays, 15 to 80 percent slopes GA75D Cabbart-Delpoint-Cabbart, wooded., loams, 4 to 35 percent slopes GA75E Cabbart, wooded-Blacksheep-Delpoint complex, 6 to 60 percent slopes GA353D Cabbart-Yawdim complex, 4 to 15 percent slopes GA372E Lonna-Cambeth-Cabbart silt loams, 12 to 25 percent slopes GA1017F Busby-Twilight-Cabbart association, wooded, 8 to 35 percent slopes

GROUP 4. MISSOURI RIVER AND TRIBUTARY STREAMS: SLOPE TOES, COLLUVIUM AND DRY TERRACES (above floodplain) 21 4.1 Colluvium and Toe Slope Shrublands ORIGINAL VEGETATION: Shrub-dominated communities along upper margins of the Missouri River bottomland and upper margins of small stream floodplains. Because of the varying minimum size of soil polygons as mapped in different counties, most of the bodies of toe slope colluvium are too small to appear on soil maps as distinct soil series except in a few spots along the Missouri River bottomlands where slopes are steep and high enough to have deposited deep fans and piles of colluvium at their toes. They are common along slope toes where adjacent slopes are steep, but typically occur in a narrow band.

CT: Sarcobatus vermiculatus-Artemisa tridentata-mixed dry prairie grasses and forbs (drier upper zone of colluvium) CT: Artemisia cana-diverse mesophytic grasses and forbs (on more mesic and finer textured material at slope toes)

Figure 13. Silver sage (Artemisia cana ) on colluvium below toes of steep slopes, Kipp BLM campground. Bands of colluvial shrubland are abundant along the bases of steep slopes of the river and its larger tributaries but are seldom large enough to appear on the six county soil maps. The higher, drier greasewood-big sagebrush zone can be seen to the right, just above where the silver sage ends.

SOIL MAP CODES and SOIL SERIES: FE142 FE Kobar silty clay loam, gullied, 2 to 25 percent slopes FE278 Yamac loam, 2-8% FE279 Yamac (loam)-Delpoint (loam)-Yawdim (clay loam) complex PH792C Yamacall complex (Yamacall calcareous 55%, Yamacall 35%) , 2 to 8% colluvial slopes (Missouri River bottomland all flooded inGarfield) GA37C Abor-Marvan silty clays, 2 to 8 percent slopes

4.2 High River and Stream Terrace Shrubland and Prairie 22 ORIGINAL VEGETATION: This type occurs on lower hill slopes and with rounded and eroded remnants of higher river terraces, along the upper edge of the Missouri River bottomland and major tributaries. These include the terraces dry enough to support prairie dogs but mostly too high above the water table to support cottonwoods. Being partly sheltered from fire, this is one of the major natural habitats for big sagebrush in a landscape otherwise too flammable or too wooded for sagebrush. Where they occur along the few perennial creeks some of these dry terraces would have supported cottonwoods in the original landscape where permanent beaver ecosystems maintained the water table several feet higher than at present.

CT: Artemisia tridentata-Dry prairie grasses and forbs. SOIL MAP CODES and SOIL SERIES FE140 FE Kobar silty clay loam, 0 to 2 percent slopes FE141 FE Kobar silty clay loam, 2 to 8 percent slopes FE166 FE Marvan silty clay, 0 to 2 percent slopes FE252 FE Vanda clay, 0 to 8 percent slopes FE253 FE Vanda-Nobe clays, 0-4% slopes PE59 Marvan silty clay, 1-8% slopes PE60 Marvan-Vanda silty clays, 0-8% slopes PH30A Marvan clay, 0 to 2% slopes - Fine, montmorillonitic, frigid, Sodic Haplusterts PH94D Busby fine sandy loam, 8-15% (1 polygon) PH601A Havre (35%)-Harlake (30%)-Glendive (25%) complex, 0 to 2% slopes PH901A Lallie clay loam, 0 to 1% slopes SM116 – high terrace, ponding long but flooding rare (none mappable in VA, most flooded by Fort Peck Lake) VA76 Vaeda silty clay, alluvial fans and low terraces, acidic, (& one polygon on a ridgetop) MC50 Creed loam, 0 to 8 % slopes MC114 Marvan clay, 0 to 8 % slopes MC177 Cabbart-Havre loams, 0 to 35 percent slopes GA30B Brushton silt loam, 0 to 4 percent slopes (Argiustolls) GA39C Creed loam, 2 to 8 percent slopes GA47B Ethridge silty clay loam, 0 to 4 percent slopes (Aridic Argiborolls) GA47C Ethridge silty clay loam, 4 to 8 percent slopes GA48B Eapa loam, 0 to 4 percent slopes (Aridic Haplustolls) GA48C Eapa loam, 4 to 8 percent slopes GA66B Kobase silty clay loam, 0 to 4 percent slopes GA66C Kobase silty clay loam, 4 to 8 percent slopes GA67C Kremlin loam, 2-8% (Aridic Haplustolls) GA98B Yamacall loam, 0 to 4 percent slopes GA98D Yamacall loam, 8 to 15 percent slopes GA373D Cambeth, noncalcareous-Megonot complex, 8 to 15 percent slopes GA385B Chinook fine sandy loam, 0 to4 percent slopes (Aridic Haplustolls) GA385C Chinook fine sandy loam, 4 to 8 percent slopes (Big Dry Cr.) GA471C Ethridge loam, 4 to 8 percent slopes GA481A Eapa loam, 0 to 2 percent slopes GA481C Eapa loam, 2 to 8 percent slopes GA558C Gerdrum clay loam, 0 to 8 percent slopes GA559C Gerdrum-Absher clay loams, 0 to 8 percent slopes GA661C Kobase-Sonnett, thin surface, silty clay loams, 2 to 8 percent slopes GA731A Marias silty clay, 0 to 2 percent slopes GA731C Marias silty clay, 2 to 8 percent slopes GA861C Archin-Gerdrum loams, 2 to 8 percent slopes (mesic prairie?) GA901C Sonnett-Sonnett, thin surface, complex, 2 to 8 percent slopes GA914D Yetull-Busby complex, 2 to 15 percent slopes 23

GROUP 5. SALINE DRY DRAINS 5.1 Saline Dry Drains with seasonal flooding events ORIGINAL VEGETATION: CT: Mixed Atriplex spp-mixed graminoids and forbs adapted to high pH and sodium. SOIL MAP CODES and SOIL SERIES VA2 Aquic Ustifluvents, saline, clay & clay loams, frequently flooded, MWD & SPD, persistent water table at 40-60” MC144 Typic Fluvaquents, saline

GROUP 6. SMALL STREAM AND UPLAND WETLANDS 6.1 Intermittently Flooded Lakes, Marshes and Vernal Pools ORIGINAL VEGETATION: CT: Mixed wheatgrass spp.-

Figure 14. Dry Lake at head of UL Bend, Phillips County. Soil unit PH930A.

SOIL MAP CODES and SOIL SERIES PH28A Nishon clay loam 0-2% in closed depressions SM 121 ponding: long. PH29A McKenzie clay, 0 to 2% slopes – in closed depressions, PD, ponding: long PH93A Bowdoin clay, 0 to 2% slopes (Fourchette Cr., Telegraph Cr.) PH930A Harlake (50%)-Lostriver (40%) clays, 0 to 2% slopes (Dry Lake, Hawley Flat)

6.2 Wooded Wetland Ravines and Drains ORIGINAL VEGETATION:

Figure 15. Wooded drain on road between hwy 24 and McGuire Recreation Area, McCone County. The dominant tree is green ash (Fraxinus pennsylvanica) and there is a tall, gray-green band of buffalo berry (Shepherdia argentea) along the margin. This site had the most and healthiest Shepherdia I saw in the CMR region.

CT: Fraxinus pensylvanica/Shepherdia argentea/diverse mesic bottomland graminoids and forbs SOIL MAP CODES and SOIL SERIES: (most sites too narrow to appear on soil maps) MC4 Aeric Fluvaquents, sandy loam-silty clay loam, occasionally flooded, 0-2% MC147 Ustic Torriorthents-Ustic Torrifluvents

6.3 Small Stream Wetland Mosaic Structured by Beaver and Fire ORIGINAL VEGETATION: See the discussion of beaver wetlands in the fire frequency section. There may have been many miles of permanent wetlands maintained by beaver in all the small streams with enough continuous or seasonal flow to support them. The extent of early beaver trapping represented in the historical literature from 1835 to 1882, suggest that beaver were abundant on at least the streams heading in the Little Rocky Mountains, the Judiths and beyond in the case of Musselshell River). CT: Salix exigua-Typha latifolia CT: Scattered small Populus deltoides/diverse emergent wetland graminoids and forbs FE118 in the Armells Creek bottom (Havre and Harlem soils, occasionally flooded) PE43 Harlem silty clay, 0-2% slopes, occ flooded (only Sacajawea River) PE48 Havre loam, 0-2% slopes (higher phase than 49) (Musselshell R.) PE49 Havre loam, 0-2% slopes, occ flooded (Musselshell R.)

25 PE51 Havre-Glendive, 0-2% slopes, occ flooded GA7A Riverwash NOTE: GA7A is natural only in Big Dry Creek above the lake flood pool. Where it occurs in the mouth of Musselshell River it represents recent deposits on the previously flooded lake bottom before recent drawdowns. GA612A Havre loam, 0 to 2 percent slopes, frequently flooded GA815A Rivra Complex, (one polygon, may not appear on GIS layer)

GROUP 7. MISSOURI RIVER AND TRIBUTARY STREAMS FLOODPLAIN COMPLEX 7.1 Low Terrace Cottonwood Flats ORIGINAL VEGETATION: Cottonwoods are on the next lower terraces below 4.2 High River and Stream Terrace Shrubland and Prairie. The typical flats in 7.1 seem low enough for cottonwoods to reach the water table. They are mostly unsuitable for prairie dogs for three likely reasons: first they are subject to flooding (occasional or rare), second the soils are soft, making it easy for predators to dig them out and third because moisture availability leads to sometimes heavy growth of cottonwood and other small trees and shrubs, eliminating the open landscape prairie dogs seem to require for protection from predators. Cottonwoods seem to be absent from places were the roots can’t tap a local water table and so are absent from most sites higher than about 2 meters above the river or above small stream bottom (or stock pond margins). The Missouri River bottomland cottonwood flats in Garfield, McCone & Valley counties are all inundated by Fort Peck Lake. See Schultz (1902) for an account of the original river rapids and bottomlands of this section before flooding.

Figure 16. Wet phase of cottonwood flats at Kipp BLM recreation area. Soil is moist but rarely flooded and then only for short periods. Older cottonwoods had fire scars on the lower trunks and ice scars higher up. Only three species were found in a 1/10 hectare plot. CT: Populus deltoides/Rosa woodsii-Cornus sericea (stolonifera)

Figure 17. High phase of cottonwood flats. As the river cuts lower, old cottonwoods are still able to reach the water table with deep roots. Once they are killed by fire, however, they seem to be replaced on the higher flats lands by prairie and shrubs, while any new recruitment of cottonwood occurs on soils lower and closer to the water table. Missouri River upstream from the refuge, between Grand Island and Cow Creek.

Schultz described the vegetation of Grand Island in 1901 (p 52): “At its upper end there is a magnificent grove of tall old cottonwoods and a growth of smaller timber completely belts it [succession by a younger stand after flood scouring or new deposition of soil], The rest of the island is a level plain, covered with buck brush [Symphoricarpos occidentalis] and tall grasses.”

CT: Populus deltoides/Symphoricarpus occidentalis/Nasella viridulus-Pascopyron (Agropyron) smithii-Poa compressa

SOIL MAP CODES and SOIL SERIES: FE115 Harlem silty clay loam, 0-2% slopes FE116 Havre loam FE117 Havre silty clay loam (none in PE: all inundated or silted up by Ft. Peck Lake). The Harlem and Harve soils are all in 6.2) PH60A Havre loam, 0 to 2% slopes PH90A Harlake clay, 0 to 2% slopes PH811A Glendive (60%)-Havre (30%) loams, 0 to 2% slopes. Mature cottonwood flats VA25 Havre silty clay loam (below dam: all above dam deeply flooded) VA26 Havre-Glendive complex, 0-2% (below dam: all above dam deeply flooded) MC74 Glendive loam, protected (downstream from Fort Peck Dam) 27 MC76 Glendive silty clay loam (downstream from Fort Peck Dam) MC80 Harlem silty clay (downstream from Fort Peck Dam) MC81 Harlem silty clay, protected (downstream from Fort Peck Dam) MC82 Havre silt loam (downstream from Fort Peck Dam) MC85 Havre silty clay loam, protected (downstream from Fort Peck Dam) MC118 Pendroy clay (downstream from Fort Peck Dam) PE84 Vaeda silty clay, 0-6% slopes Musselshell River, Sacagawea River) PE85 Vanda silty clay, 0-4% slopes Musselshell River, Sacagawea River) MC106 Lonna-Havre-Glendive complex, 0 to 2 % slopes GA4A Ismay silty clay loam, 0 to 2 percent slopes, occasionally flooded GA59A Hanly loamy fine sand, 0 to 2 percent slopes, rarely flooded (see GA591A) GA61A Havre loam, 0 to 2 percent slopes, rarely flooded GA561A Glendive fine sandy loam, 0 to 2 percent slopes, rarely flooded GA562A Glendive fine sandy loam, 0 to 2 percent slopes, occasionally flooded GA563A Glendive-Havre complex, 0 to 2 percent slopes, occasionally flooded GA564A Glendive loam, 0 to 2 percent slopes, rarely flooded (GA591A Hanley loamy fine sand, 0-2%, occasionally flooded (as opposed to rarely flooded in 59A)) GA611A Havre loam, 0 to 2 percent slopes, occasionally flooded GA617A Havre, Harlake, and Glendive soils, channeled GA619A Havre-Glendive complex, 0 to 2 percent slopes, occasionally flooded (Musselshell R.) GA703A Lonna-Havre-Glendive complex, 0 to 2 percent slopes (GA851A Rivra Complex is listed in 9.2 below as a disturbance artifact of flooding since it occurs at tha mouth of Big Dry Creek into Big Dry Arm. It would have been 7.1 cottonwood flats before impoundment)

7.2 River and Small Stream Wetlands (moist soils) ORIGINAL VEGETATION: This type occurred more extensively along the Missouri River in the pre- dam landscape because heavy flooding accompanied by massive ice jams as described by Schultz (1902), scoured the bottomlands during spring ice-melt. This reset the process in the lower parts of the bottomlands led to new primary succession of wetland forbs, followed quickly by seedlings of willow ands cottonwood. James Schultz (1902 p. 57) described dynamic erosion by the unrestrained river even in September, and the woody succession that took place:

“ In one day the ever-shifting channel will remove all traces of a long, wide bar or island several feet in height. Often, as we rowed or sailed along, we could see them melting away, yards and yards at a time, and great chunks of the bottom, ten, twenty, even thirty feet in height, were continually falling in with a resounding splash. The careful navigator will do well to keep out from the cut banks. Where a bottom wears away, the bottom on the opposite side fills out, and at a rate which can be accurately measured by the growth of the trees. Always at the outer edge are cottonwood and willow sprouts; back of them belt after belt of timber, each one larger than the other by a year’s growth, until finally one comes to the full-grown trees, tall, rough-barked and wide of girth.”

SOIL MAP CODES and SOIL SERIES: FE118 Havre and Harlem soils, 0-2% slopes, occasionally flooded (only the polygonss of FE118 in the MO River bottomland. Others are in 6.3 beaver wetlands and in 2.5 Mesic prairie in deep creek & coulee valleys like CK Creek & Rock Creek. PH601A Havre (35%)-Harlake (30%)-Glendive (25%) complex, 0 to 2% slopes (none in VA because all flooded by Fort Peck Lake) MC143 Typic Fluvaquents, frequently flooded MC176 Havre-Bigsandy loams, 0 to 2 percent slopes, frequently flooded CT: Salix exigua-diverse early succession forbs 28

GROUP 8. WATER 8. Water

GROUP 9. DISTURBED SOILS 9.1 Fort Peck Dam, other disturbed soils SOIL MAP CODES and SOIL SERIES: VA46 Phillips loam, 0-5% slopes (graded surfaces below Fort Peck Dam: not really Phillips loam) VA78 Ustorthents (not on printed VA map) MC “DAM”

9.2 Former bed of Fort Peck Lake PE73 Riverwash, freq flooded (no soil taxonomy, would have been mostly Harlem & Havre in the original landscape before the lake flooded & silted up the lower end of Musselshell River) GA7A “Riverwash” = silted up lower Musselshell River as in PE73. GA7A, however, is natural in the unflooded channel of Big Dry Creek where slightly dryer portions of it would have been cottonwood flats while wetter parts would have been beaver wetlands. (GA851A Rivra complex, 0-2%, frequently flooded (on new soil map, at mouth of Big Dry Creek, another likely artifact of Fort Peck Lake impoundment since water comes right up to it). Portions would have been 7.1 cottonwood flats before impoundment.

GROUP 10. (none actually on the refuge?) 10. “denied access” There was only a tiny part of the refuge, in Garfield County, not mapped because access was denied.

APPENDIX 2. CHRONOLOGY OF HISTORICAL EVENTS RELATED TO FIRE AND NATIVE SPECIES AT CMR, 1730-1970

Note: this is a list of selected events of historical interest in the CMR region. It is by no means complete but may be of use to staff and future researchers at CMR as background for further work. See Frost 2008 report on pre-European fire frequency for bibliography of sources.

1519 Cortes brought horses to Mexico.

1598 Don Juan de Oñate took soldiers, missionaries, seeds and over 1500 head of horses and mules to New Mexico to conquer the Pueblo Indians.

1608 Apache and Navaho Indians drove Spaniards out of their early settlement in New Mexico, seizing livestock, including horses. Indians established herds of Spanish horses and traded them to other tribes.

1632 Fray Benavides in his journals notes the war chief of the Gila Apaches riding a horse, the earliest such report. See http://www.redoaktree.org/indianhorse/history2.htm for an outline of early expansion of use of the horse by southwestern Indians.

1673 Jolliet and Marquette discover the mouth of the Missouri River.

1680 Pueblo Rebellion – all the Spanish were driven out of New Mexico by the Tewe and Navaho. Over the next several decades the large number of horses left behind by the Spanish were traded across the west, and by the 1730’s nearly every tribe of the southwest and of the plains up to southern Canada had them.

1737 La Verendrye reached the Mandan villages of central North Dakota. The Mandans did not yet have horses and were still using fire to burn the prairies to attract buffalo to near the villages where they could be hunted on foot. They also used fire for circle hunts but since they had fixed villages and grew corn, the Mandans were not as dependent upon the buffalo as the Blackfeet and other plains Indians became after they acquired horses.

1730-1740 Blackfeet (and other people sof the CMR region) obtained horses (see Ewers, 1958, Schwartz 1973) so they only had been equestrian for around 140 years by the time James Schultz began living with them in 1877.

~1750 The Mandans acquired horses.

1750-1850 Some minor reduction in buffalo numbers might have been likely as Blackfeet and the other Indians of the CMR region became efficient in using horses to kill them. With a slight reduction in grazing pressure, there could have been a corresponding slight increase in grass cover on the shortgrass prairie and a resulting fire frequency increase. If so, however, this effect may have been cancelled out by reduced use of fire for attracting and hunting bison by First Nations peoples. The reduction in fire frequency after the early 1700’s seen in the fire scar chronologies at CMR suggests that this was the case.

1787 A large war party of Piegans (Blackfeet) brought home horses abandoned by a Spanish column in fear of attack by the Indians, around latitude 32 degrees. One of several accounts of long exploratory trips by the Blackfeet, this would put the encounter somewhere south of Albuquerque, New Mexico (Ewers p. 196).

Late 1700’s The Blackfeet began to trade with the Hudson’s Bay Company at some of the seven forts built by the Company from 1786 to 1808 along the North Saskatchewan River in present day Alberta, Canada. The Blackfeet acquired guns there and began to use them in buffalo hunting, although there may have been little impact on buffalo with primitive early guns: as late as 1881 some Indians still preferred the bow and arrow. Blackfeet had guns before others because of their trading location between the Hudson Bay Company in Canada and the other tribes to their south. The Blackfeet also had a commercial advantage by having a variety of British Canadian goods to trade with southern groups such as the Crow and Flatheads.

Late 1700’s The Hudson Bay Company published an annual list of fur species that they would buy. Beaver headed the list, wolf pelts were desired and buffalo robes were last. The trade was largely via their posts on the North Saskatchewan River and the long transport route to the east, including portages, made bulky bison robes unprofitable. This changed after 1832 when the American Fur Company began sending an annual steamboat up the Missouri River to Fort Union on the spring rise in the river to collect the results of the winter’s trade with the Indians.

1803 With the Louisiana Purchase, Thomas Jefferson added the whole watershed of the Missouri to the United States.

1804-1806 Louis and Clark spent the winter from October 1804 to April 1805 at the Mandan and Hidatsa villages of central North Dakota before pushing on up the Missouri through CMR in 1805. Captain Louis returned downriver through CMR in 1806 while Clark explored down the Yellowstone. The Mandans, who had horses for some 50 years by then, were using them to hunt buffalo, but since they were farming and living in fixed villages, they were still using fire to attract them:

Dec. 7, 1804: Capt. Lewis took 15 men & went out to join the Indians, who were at the time he got up, Killing the Buffalow on Horseback with arrows which they done with great dexterity ... ____William Clark, December 7, 1804, written at Fort Mandan.

“The Plains are on fire in view of the fort on both Sides of the River, it is Said to be common for the Indians to burn the Plains near their villages every Spring for the benifit of their horse, and to induce the Buffalow to come near them.” ____William Clark March 30, 1805, written at Fort Mandan.

1807 Manuel Lisa explored the Yellowstone as far upriver as the mouth of the Big Horn river and established the first outpost, Fort Manuel there.

1808 The American Fur Company was organized by John Jacob Astor but initiation of activities was delayed by the War of 1812.

1809 The Missouri Fur Company sent 350 men on a trapping expedition. Some reached the three forks of the Missouri in 1810 and established a post there but it was short lived because of the Blackfeet. The company dealt in furs in the region until around 1816. There appears to have been 23 years of sporadic but poorly documented trapping on the upper Missouri, including CMR, by free trappers beginning around this time until around 1833 when better records are available.

1822 The Rocky Mountain Fur Company was organized by William H. Ashley and Andrew Henry and initiated trapping for beaver in the Three Forks area of Montana (on the tributaries of Madison, Jefferson, Gallatin Rivers, in the lands between Helena and Bozeman) for several years. Company ownership succeeded to Jebediah S. Smith, David E. Jackson and William L. Sublette in 1826. The company traded large numbers of beaver and other furs profitably for 12 years until its dissolution in 1834 after intensified competition with the American Fur Company.

1822 The American Fur Company began trapping on the lower Missouri.

1827 The American Fur Company bought out the Columbia Fur Company and moved into trapping on the upper Missouri

1829-1831 Fort Union was built by Kenneth McKenzie of the American Fur Company on the Montana-North Dakota line. Free trappers from Fort Union ranged at least as far upriver as the Milk River and lower CMR, bringing beaver pelts back to this fort to sell. This was for many years the headquarters of the American Fur Company on the upper Missouri and Charles Larpenteur (see “Forty Years a Fur Trader on the Upper Missouri”) was employed here for many years. Schultz 1937 and Ewers 1958 have maps with dates of establishment of the forts on the upper Missouri.

Early 1800’s An old Indian friend recalled to James Shultz (1962), from sometime before 1825, a description of the last buffalo jump used by their particular band of Piegans. This was still being done even though they had had horses for 60 or 70 years. When he was only 8 years old he participated in the last use of this method of killing buffalo. Some of the older individuals in his band had participated in the first acquisition of horses from the southwest some 80 years earlier. He remembered all that they told him and passed it on to Schultz in the early 1900’s so this was an actual oral history of the acquisition of horses by the Blackfeet.

~1830 beginning of decline of wolves in Montana (when did poisoning for pelts begin?) when did poisoning of wolves and coyotes become significant? The use of poison could mark the early starting point of the prairie dog increase reported in the historical literature from several states.

1829-1831 James Kipp of the American Fur Company built Fort Piegan at the mouth of the Marias, initiating trade with the Piegans. Fort was burned and rebuilt in 1832 as Fort McKenzie (p 97). The initial emphasis was on inducing the Indians to trap beaver and traps were lent them for that purpose.

1832 The Yellowstone was the first steamer to reach the upper Missouri River, carrying George Catlin to the Mandans at Fort Clark (Bismarck, ND). He traveled on to the newly built Fort Union near the Montana line. After this a steamer came regularly with the rise of the river in June of each year to bring down the furs and buffalo robes collected during the winter’s trade with the Indians. The steamer names and arrival dates are often given in Larpenteur.

1833 Two American Fur Company steamers, the Yellowstone and the Assiniboine reached Fort Clark near Pierre, SD together Union together. The Yellowstone took on a load of 7000 buffalo hides and returned to St. Louis. The Assiniboine continued on to Fort Union, the head of steam navigation at that time.

1833 Prince Maximillian of Wied travelled on the Yellowstone to Fort Union, then by a cordelled keelboat up through CMR to Fort McKensie (near Fort Benton today) where he spent 2 months that summer (see his “Travels” book). Karl Bodmer who accompanied him did his famous paintings of the upper Missouri River at the same time, apparently only one of which, showing the “white castles” formation now inundated by Fort Peck Lake, was painted in CMR. A second painting, showing a file of buffalo coming down a ravine to the river to drink was painted somewhere between Cow Creek and Grand Island, just above or on the upper end of CMR. Maximillian’s daily journals have some descriptions of CMR while passing through up and downstream and much information on the fur trade.

1833 Prince Maximillian was shown a skin of a moose killed on the Milk River, one of a number of records from that site beginning with that of Lewis and Clark in 1805.

1833 General Ashley’s beaver trapping expedition to Pierre’s Hole, Idaho west of Yellowstone trapped 1000 lbs of beaver skins in 7 or 8 days (Larpenteur p. 28). By late August, they left with 30 packs or 1800 beaver making 3000 lbs on the navigable part of the Yellowstone to take down to the mouth of the river at the Missouri on buffalo skin boats (p. 29). Then on to St. Louis (Larpenteur 1898). Ashley began trapping after the War of 1812 and some of this was on the upper Missouri.

1833 Charles Larpenteur hired by American Fur Company at Fort Union, began to keep his journals later condensed into “Forty years a fur trader on the upper Missouri”, Larpenteur 1872 [1989]. During his time some 26 forts or trading posts were built along the upper Missouri from Fort Vermillion near the Iowa line to the Three Forks area south of Helena. Often there were two in the same vicinity built by opposing trading companies. When a company left off trade at one post they often burned it to prevent its use by another company.

1833 Keel boats, like that of Lewis and Clark, hauled upstream by men on shore with tow ropes called cordelles, were still being used to ascend the Missouri but were about to be replaced in deepwater parts of the river by steamboats. Steamers were frequent as far as Lexington, MO.

1833 about 25,000 beaver skins already being taken annually out of the upper Missouri by the American Fur Company (Maximillian, notes on June 24, 1833)

1833 Apparently no buffalo left in eastern Nebraska but plenty of buffalo around Laramie, WY (noted by Larpenteur while heading west on the overland expedition to Pierre’s Hole).

1833 Some 5 or 6 cattle, including 2 milk cows were established on the Montana, North Dakota line after being taken by Larpenteur and others on the two to three thousand mile beaver trapping expedition to Pierre’s Hole, Idaho (Larpenteur p. 35, 39). These founded a small herd that was kept there along with a small hay field for several decades and are mentioned from time to time in Larpenteur. These may represent the first permanent cattle in Montana (or on its border).

1833 Fort Cass built on the Yellowstone about 2 miles below the mouth of Big Horn River. First trade was in deer and elk skins (Larpenteur, p. 28).

Fall 1833 beginning of trade in buffalo robes from Montana down the Missouri via the annual steamboat. New Fort William was built at the mouth of the Yellowstone on the Montana line in competition with Fort Union. Assinboins came with 200 buffalo robes to trade. This event and year mark the initiation of serious commercial exploitation of the buffalo in Montana, which reached its culmination only 49 years later with their extirpation in the winter of 1881-1882.

1833-1882 Despite government opposition, liquor was always the principal and most profitable article of trade with the Indians (Larpenteur p. 46).

1833 A very warm fall kept buffalo from migrating south from Saskatchewan as far as Fort Union and Fort William (only 2 ½ miles apart) and this was true all along the Missouri (see also the warm winter without buffalo in 1862). The only meat was an occasional deer or elk killed in the vicinity. p 47-50). Cottonwood logs were used for the stockade. June 24, 1834, annual spring shipment: only 70 packs of robes (10 x 70 = only 700 buffalo robes brought in all winter, apparently transported by Indians from some distance), 16 packs of wolves (x 30 = 480 wolf skins and a few fox skins. No beaver mentioned, probably all cleaned out locally. The same nonappearance of buffalo happened again in the warm winter of 1862-1863 (see below).

1834 Usually only one steamer a year as far up as Fort Union, usually in June. In 1834 it was the Assiniboine.

1934 Hay making noted at Fort Union (Larpenteur p. 58), free trappers p. 63. June 1836, the steamboat arrived again p. 80.

1835 Buffalo returned the winter of 1834-1835, but were driven far enough from the fort over the winter that hunters went after them and camped, sending back meat on horses.

1835, July: Gardipee and the Deschamps family went on a beaver hunt, leaving from Fort William (opposite Fort Union). In that Fall 2 other trappers left to trap beaver on the Milk River, which “abounded with beaver” but were killed by the Blackfeet. p 76. a herd of 300 bison was seen near the fort.

1835-1850 Phase 1 of dewatering of many small streams on the upper Missouri, from tributaries of the Milk River, through CMR at Fort Benton began with trapping of beaver and eventual

collapse of beaver ecosystems and entrenching of deep channels in their former bottoms. Phase two began with diversion of small stream headwaters around the Judith Mountains for agriculture around 1880-1890.

1836 The Blackfeet were described as “very rich in horses” (Larpenteur p. 76) November, a free trapper canoed down the Missouri to Fort Union with a pack (60 skins, 100 lbs) of beaver skins worth about $500.

1836 The American Fur Company trade was still mostly in small furs on the upper Mississippi p. 91. Hudson’s Bay Company trade limited to “fine furs” because of the difficulty of portaging heavy buffalo robes from the Saskatchewan River are north of CMR back east.

Late June, 1837 The annual steamer arrived as usual, bringing smallpox and initiating the 1837 smallpox epidemic among the Indians at Fort Union where their numbers were reduced by half by spring 1838. Indians continued to bring in robes for trade even during the epidemic.

1838 Still usually only one steamer per year, the American Fur Co’s Antelope (Larpenteur p 121).

1838 Buffalo were reported to be becoming very scarce in North Dakota as far down as the Mandans p. 127. Larpenteur’s only mention of prairie dogs in his entire book (he killed some while hunting).

1838-1839 Spring shipment from Fort Union: 250 packs (2500) buffalo robes plus many small furs. To St. Louis in 8 Mackinaw boats (figures do not cover other companies or other posts’ trade but it should be possible to find data from them to get an estimate of the numbers being taken from Montana annually).

1840 The steamer “Trapper” from St Louis to Ft Union March 31 to June 27 (nearly a three month trip). Larpenteur p 137.

1842 Bison still abundant in all of eastern Montana from the mouth of the Yellowstone River up to Rosebud River. Larpenteur built Fort Alexander on the south side of Yellowstone River between the Rosebud River and the Bighorn.

1843 Fort Chardon built at mouth of Judith River just upstream from CMR.

1844 Jim Bridger with 30 men trapped on the Milk River, was disappointed in the number of beaver, but other trappers had been working it for years and it was likely trapped out. Beaver there were said to be abundant by trappers working the river in 1835. The American Fur Company at Fort McKenzie upstream had been lending steel traps for beaver to the Indians since around 1732 and unknown numbers of “free trappers” may have worked the Milk River area sporadically before that. There were milk cows at Fort McKenzie at the mouth of the Marias River (near Benton).

1845 Commencement of the meat trade at Fort Union. p. 199. In 1846 meat was scarce there.

1845-1846 Larpenteur in charge of new fort on Poplar River between Milk River and Ft Union. In winter 1846 he bought only 35 packs = 350 buffalo robes and a few small furs.

1846 Fort Benton was built by the American Fur Company near the site of the earlier Fort McKenzie, just above the mouth of the Marias River. p. 225. “the steamer, which never failed”, arrived at Fort Union, Larpenteur took supplies from there in a keelboat through CMR to Fort Benton, because the river was low in July.

1846 Description of cordelling a keelboat up through CMR to Fort Louis (near Benton). Took 70 days & all Larpenteur had to say about CMR was about killing 35 deer and 15 elk on the way (there might be more in his original journals, some of which still exist)(Larpenteur. p 210).

1847 Hay field at Fort Union was still in operation after 13 years (Larpenteur p. 220) when Indians attacked and killed 4 oxen there.

1848 Larpenteur travelled up the Missouri from Fort Benton to Sun River. Killed 3 grizzly bears p 227. “When one gets to the mountains he is out of the range of game” (Larpenteur p. 229).

1859-1860 Bill Featherland, an employee of the American Fur Company camped in CMR on Featherland Island about four miles below the mouth of Hell Creek (the island now drowned by Fort Peck Lake) and “killed, poisoned and trapped 1500 wolves, to say nothing of coyotes and kit foxes” in a single winter (Schultz p. 131).

1860 Cattle were a regular part of any trading outfit (Larpenteur. p. 265). Once they had helped haul in goods they often remained in central Montana to become breeding stock.

1860-1861 (winter) Larpenteur at his post near Poplar River, downstream from CMR, traded 2000 buffalo robes + small skins p 273.

June 1861 Two steamers at Fort Union, the Chippewa and the Spread Eagle. The Chippewa would have continued up to Poplar River but was burned by a drunk 6 miles downstream. Larpenteur sold his cattle (never said how many) and wagons to some men taking them to a ranch in the Bitterroot Valley (near Missoula p. 276-278).

1861-1862 Severe winter at Fort Stewart near mouth of Poplar River “Buffalo were so plentiful here last summer that they ate up all the grass; it looked as though fire had burned the prairies.” 20 head of cattle died at Fort Union that winter, the whole herd) . Larpenteur nearly starved but bought some 1450 buffalo robes.

1862 Two steamers arrived at Fort Stewart in June, the Shreveport & the Emilie p 288. Now within CMR, the site for Fort Galpin (near Fort Peck) was selected at Moose Point, 10 miles up from Milk River, (this was still downstream from where Louis & Clark saw moose: their sighting was 10 miles up from Big Dry Creek p 289. The exact location for Maximillian’s 1833 moose skin from the Milk River is not known.

Aug 1862 water so low the steamer Shreveport had to stop at Milk River.

1862 James Fergus came to Montana with the Captain Fisk Expedition, established a ranch near Helena and began to build a herd of cattle and horses that he later moved to the head of Armells Creek in 1880, a year ahead of the final extirpation of the buffalo.

1862-1863 Another mild winter: the annual buffalo migration from Canada again failed and there were with no buffalo along the Missouri from somewhere in CMR down to Fort Galpin/Fort Peck and beyond. Indians & traders at the fort were starving (see the similar events in the warm winter of 1833).

Spring 1863 Three steamers, the Robert Campbell, the Nellie Rogers and the Shreveport came to Fort Union, at least two of which went up to Fort Benton on the spring flood. With low water goods had to be offloaded at Milk River and taken up to Fort Benton overland by wagons drawn by oxen (Larpenteur p.298).

1864 Steamer Benton (large) to Fort Union with 50 tons of Army commissary freight and 150 men. 2nd steamer Yellowstone passed through in June on way to Fort Benton. 300 buffalo were near the fort (p. 302-303). The Army had 5 steamers to go up to Union on way to Yellowstone to fight Indians (Indians stole all their horses p. 306).

Spring 1865 Steamer Yellowstone to Ft Union. American Fur Company sold out to North West Fur Company including Fort Union. Steamer Hattie Mae arrived at Fort Union I September. p. 314.

Winter 1867-1868 Larpenteur traded 2000 buffalo, 1800 deerskins, 1000 wolves and 900 elk at Fort Union, a typical good winter’s trade for one fort.

1870 The Baker massacre – the US Army killed over a hundred friendly Piegan Blackfoot men, women and children at a camp on the Marias River as punishment for a raid by a different band.

1876 The battle of the Bighorn (Custer’s last stand) spurred more Army involvement and efforts by some legislators to provoke intentional elimination of buffalo in order to break Native American resistance to white encroachment. In the second-half of the 19th century European buffalo hunters, armed with powerful, long-range rifles, began killing the animal in large numbers. Individual hunters could kill 250 buffalo a day. By the 1880s over 5,000 hunters and skinners were involved in this trade. It is claimed that the killing of buffalo was supported by the U.S. military in order to undermine the survival of the Plains Indians. "Let them kill, skin, and sell until the buffalo is exterminated, as it is the only way to bring lasting peace and allow civilization to advance." - General Philip Sheridan

1877 James Schultz moved to the upper Missouri, married a Blackfeet woman Natahki, just in time to chronicle the last five years of the trade in buffalo robes, and the lives of the Indians dependent upon following the buffalo. Much of this time he spent buying buffalo robes and

other skins at Carroll on the Missouri River (now part of CMR), his time at Carroll later described as “the happiest years of my life”.

1880 James Fergus began acquiring land for cattle and sheep ranching land on upper Armell’s Creek, had 8600 acres in the early 1880’s. 1880 had 1000 cattle. In 1884 he began acquiring sheep. 1896 bought Eugene Townsend’s 100 acres along with 4000 sheep, 60 horses and 125 cattle. By 1892 Fergus and his son Andrew had 12,200 head of livestock (2000 cattle, 9000 sheep and 1200 horses). I found no mention of sheep around early trading posts that did have cattle and occasionally a few pigs and goats.

1880 F.J. Haynes of Fargo, ND traveled up the Missouri River taking the first photos of CMR (see his Cow Creek photos in the fire history section of this report [Frost2008]) and on this and other trips including the Yellowstone River area, took stereo photos of the eastern prairies. He was followed later in the CMR region by another photographer, L.A. Huffman. See the Montana Historical Society’s photo archives in Helena.

1881-1882 Final winter buffalo robe season, November to April. Indian populations following the last herd unknowingly fell in around the edges, completely encircling and destroying the herd during the winter.

1882 Buffalo extirpated except for small remnants in a few places such as the Yellowstone and about 15 in the CMR badlands.

1882 A traveler in Yellowstone commented that prairie dogs had increased with the demise of the buffalo and similar observations were reported with the passage of the buffalo elsewhere. Rather than anything to do with buffalo, these observations coincide with an escalating increase in trapping and poisoning of wolves and demise of all other predators that fed on the poisoned carcasses. The probability should be considered that there was a real population increase base on that cause.

1882-1910 Cattle and sheep replace buffalo (to excess).

1882-1890 Short window of opportunity for grass increase and fire frequency increase following almost instantaneous demise of buffalo in the CMR region (extirpation faster than any other region because it was the last) and intensive grazing by introduced cattle and sheep. (the increased fuel continuity shows up in all three fire scar chronologies at CMR with a series of closely spaced fires beginning in summer 1882.

1884 Steamer “Red Cloud” struck a sawyer and sunk a few miles downstream from Hell Creek.

1885-1890 Phase two dewatering of small streams such as Armells Creek began with construction of irrigation ditches in their headwaters (Phase one was the elimination of beaver- maintained wetlands that retained water). James Fergus in the late 1890’s noted that he had 30 miles of irrigation ditches on his lands in upper Armells Creek. In 1901 James Schultz commented on many of the streams of CMR as being wet, formerly wet or dry. Upstream he

observed the drying of several streams in the 21 years since he had first seen them around 1877. On leaving Fort Benton he noted: “On we went past the fort, and down over the Shonkin bar at the mouth of the stream of that name which puts in here from the Highwood Mountains to the south. It is a stream no longer. Once it was a good-sized creek of pure mountain water. Schools of trout lived in its clear depths, and the beaver bridged it with their dams. Then came the white man and used the water to irrigate vast tracts of the barren plain, so nothing now runs in the old channel but a little seepage of brown alkaline water. The trout are dead, the beavers have vanished, never to return” (Schultz 1901).

1883 James Fergus’s brother William began acquiring land. At his death in 1905 William had 8000 acres and 20,000 sheep near the Judith Mountains.

1886-1903 After the collapse of the buffalo trade James Schultz and Natahki lived on the Blackfeet reservation where they raised livestock and operated a hunting guide service. Schultz continued to record Blackfeet history for the rest of his life.

1893 James Fergus complained that wolves had killed 140 sheep out of a flock of 400 and one herder’s horse. Fergus hired “wolfers” to get rid of wolves and lobbied the state legislature for a bounty. Carroll, after the demise of the buffalo was described as a wolfers’ town.

1900 James Fergus and son had 2500 cattle, 1000 horses and 9000 sheep. He began to advertize his land for sale” On account of old age and ill-health, will sell about 10,000 acres of land in Fergus county, Montana, nearly all on tributaries of Armells creek, on the north side of the Judith mountains; has 30 miles of irrigation ditches, 80 miles of fences, nine homesteads or sets of ranch buildings, together with about 2,000 good cattle, 9,000 well bred sheep and 1,200 horses, originally bred from grandsons of Mambrino Patchen and other good stallions; 6,000 bushels of oats, several hundred tons of hay, two blacksmith shops; one-third interest in 55 miles of the phone line and some private; 200 acres in crop this spring; an interest in sheep shearing sheds and apparatus, and all the necessary implements for running such a ranch, including a post office. Fergus Livestock and Land Co. James Fergus, President, Armells, Fergus County, Montana. He died in 1902 at 87 years of age.

1870-1920 Window of prairie dog increase related to poisoning of wolves (and coincident poisoning of coyotes, eagles, bobcats and mountain lions that happened to feed on strychnined deer carcasses used for wolf bait. This phase of expansion came to an end when ranchers and government programs began regular poisoning of prairie dog towns around the 1930s.

1900-1910 Land in the CMR region reached saturation with sheep and cattle (see graph for peak numbers around 1900).

1901 James Schultz boat trip from Fort Benton to Milk River, described conditions and many events in CMR that occurred during his time there from 1877 through 1882. There was Douglas fir as far east as 4 miles downriver from Round Butte.

1901 Wolves largely extirpated (first trapping and poisoning for hides, then poisoning to protect sheep and calves) except or remote areas such as the river breaks. Along the more than 200 miles from Fort Benton to the Milk River, Schulz saw wolf tracks or heard them at night only five times and only in the stretch of river now in CMR (Schultz 1901 pp. 17, 39, 52, 100, 126).

1910-1930 Regional fire frequency very low, following removal of fuel connectivity by grazing by ever increasing numbers of cattle and close cropping of fuel grasses by sheep. Fires reduced to small isolated patches. In the breaks, fire frequency would have been reduced to only those ignited by lightning (or settlers) . With no fires coming in from the plains above, fire frequency would have been reduced. Given long intervals for accumulation of juniper and dead woody fuels, future fire intensity should be hotter.

1920s-1930’s Systematic poisoning of prairie dogs begun.

1960’s Fire suppression became effective in the CMR region.

1967 27 Rocky Mountain bighorn sheep reintroduced to Mickey Butte & Brandon Butte area. Cattle removed from U.L. Bend wilderness.

1992 Bubonic plague (Yersina pestis) introduced to US 1894-1900 by ships carrying infected rats from the 1894 global pandemic, reaching CMR region with outbreaks decimating prairie dog colonies towns beginning in 1992.

APPENDIX 3. GLOSSARY OF TERMS AND CONCEPTS OF LANDSCAPE FIRE ECOLOGY

Landscape fire ecology – the study of the interactions between fire, organisms and landscape.

Adversity habitat – marginal habitat in which a fire-dependent species may survive a period of fire exclusion: the last habitat in which a species is found before it dies out. With Venus flytrap this is the dry margin of its preferred wet habitat. With Hudsonia montana it is the thinnest soils on rock ledges where a few plants may eke out an existence while 99% of the population is overgrown and eliminated by shrubs. Area fire frequency – the fire-return interval for a large area such as a National Forest or Wildlife Refuge. This is the incidence of fire within the area, even though only parts of the area may burn, so fire frequency may be reported as multiple fires per year (compare Site fire frequency). Backing fire - fire moving slowly, against the wind (assumed rate of spread of backing fires is relatively constant regardless of wind velocity. Flames lengths are typically shorter but residence time is longer and may kill larger diameter stems of shrubs and saplings. Bay forest - fire-infrequent forest type dominated by evergreen trees like red bay, sweet bay and Gordonia, found in fire tension zones between pocosin, or other frequent-fire type, and more fire- protected vegetation. Bottlenecks – Pinch points or choke points in the landscape where fire must pass through a narrow window or corridor to reach the next fire compartment. As the number of bottlenecks increase, fire frequency in the landscape decreases. Bottlenecks that occur where the channel is aligned with the prevailing wind would be less effective in reducing fire downwind than crosswind bottlenecks where fire would usually have to flank through to get to the other side. Canebrake - wetland fire vegetation type with 50% or more cover of cne (Arundinaria). (canebrake was often called reeds, marsh or even pocosin in Colonial Era literature). Canebrake reaches its best development on shallow to medium deep peat soils and wet mineral soils of fluvial bottomlands. Canopy thinning fires – fires which kill some canopy trees but fail to initiate crown fire or to kill all trees from below. Charcoalization – preservation of flowers, seeds and other fossils by being charred by wildfires shortly before being covered by water and sediment. Charcoalized parts may be preserved for millions of years, leaving a record of ancient fires. Dry lightning strikes – lightning strikes that occur in the dry fuel outside the rain path of a small storm or at a time before a cloud begins to produce rain. Ecological fire effects – effects of fire on vegetation structure, species composition, and physical characteristics of habitat. For example, the restriction of big sagebrush (Atemisia tridentata) to naturally fire sheltered sites or regions is an ecological effect of fire. Effective windspeed – A way to account for slope effects in determining rate of fire spread. For example, midflame windspeed is adjusted upward in relation to the degree of slope for a fire moving uphill with the wind. Facultative pyrophyte - a plant species not dependent upon fire but which reaches its best development in fire communities or exhibits adaptations that permit it to survive under a natural fire regime. Fetch – The length of continuous fuel in a fire compartment possible to be traversed by fire from its upwind to its downwind side. (while fetch will vary in length according to the wind direction at

time of fire, it can be mapped in terms of the prevailing wind direction during fire season. The longer either the fetch or reach, the higher the fire frequency. Fire adaptations - characteristics like thickened bark, rapid resprouting, serotiny, specialized responses to increased light after a burn and fire-stimulated seed production. There are two suites of fire adaptations, those with characteristics that decrease frequency and intensity of fire, and those that increase frequency and intensity. Examples include production of litter which decomposes rapidly therefore decreasing flammability, or production of litter which decomposes slowly and promotes flammability. Fire behavior - variation in rate of spread and intensity of fire moving through the landscape. Rate of spread and intensity respond to changes in fuel, vegetation structure, topography, wind, and diurnal changes in temperature and humidity. Firebreak – natural landscape features which stop the flow of fire under average burning conditions. These include rivers, other bodies of water, steep slopes (downwind side), deep ravines, unvegetated sand and rock and non-pyrophytic vegetation. Firebreaks form the boundaries between fire compartments. Firebreaks may be temporary, as in a portion of landscape that has already burned and has not had time to rebuild vegetation sufficiently continuous or mature enough to carry fire again. Fire barriers are rarely absolute: under severe burning conditions fires have been known to spot across bodies of water over a mile wide. By definition, a fire barrier stops most fires that occur under average wildfire conditions. Firebreak (silvicultural) – to create a “soft break” remove woody stems and maintain road edges in easily managed natural grass or herbaceous layer, versus a “hard break”, a plow line, road, dirt, gravel or water. Fire compartment - a unit of the landscape with no internal firebreaks and nearly continuous fuels, so that an ignition in one part would be likely to burn the whole unless there were a change in weather or fuel moisture. Fire compartment size is a major driver of fire frequency. The fire compartment is one unit in mapping fire frequency: the larger the fire compartment the higher the fire frequency Fire corridor - pathway for fire flow between fire compartments (see fire path). Fire-dependent species – a species that depends upon fire to maintain or prepare its habitat, or to facilitate completion of some phase of its life cycle, such as scarification of seeds or opening of cones. More broadly, this includes any species that ultimately will become rare or go extinct without fire. This may include a large percent of the world’s flora! Fire effects - changes in things such as fuel load and vegetation structure, effects on fauna, mineral nutrients, soil composition and structure. Physical effects range from simple litter removal and herb layer reduction to shrub reduction, understory thinning, understory reduction, stand thinning or canopy destruction. See ecological fire effects. Fire effects gradient – changing distribution of plant communities across a landscape based on differential effects of fire on different community types. Fire ephemeral species – plants that persist as seeds or spores, in the seed bank, appearing in large numbers after a fire. Most are annuals, plants with seeds or spores widely dispersed by wind, that may persist for decades before fire prepares a site for germination. Examples: the “fireweeds” such as Erectites hieracifolia and Epilobium spp, and Funaria hygrometrica, a moss. Fire-exposed - referring to portions of the landscape lacking natural firebreaks or environmental factors to slow the spread of fire, especially broad flats, south slopes and ridges. Site exposure grades from fire-exposed to fire-sheltered to fire-protected. Typically, in a landscape, the most fire- exposed sites support the most fire dependent species. 2

Fire facilitator species - species with adaptations in litter, plant growth form, or plant part structure or chemistry which facilitate spread of fire or enhancement of fire intensity. Presumably these features promote their survival. Fire filter – topographic, soil or vegetation features that temporarily reduce fire intensity or rate of spread. Examples: steep north slopes, areas with complex or rugged topography where all the land is in slope, arid land soil types that produce thin or discontinuous fuels, areas disturbed by animals such as densely distributed prairie dog towns and any vegetation type with poor fuel connectivity. Fire filters reduce fire frequency in two ways. First, a certain percentage of fires will be detained long enough for rainfall or cool, moist nocturnal conditions to extinguish the fire. Second, even fires that pass through the filter will have been delayed: a certain number will not spread as far as they would have before being extinguished by rain or other events. The density of fire filters and fire barriers, along with fuel type, ignition source and fire compartment size determine the fire frequency in a region. Fire flow - the movement of fire over the landscape. Characterized by rate of spread and relation to wind direction (backing fire, flanking fire and headfire). Flow is sustained by fuels such as grass in prairie and savannas, and needle or leaf litter in forests, Fire frequency – the average time between fires sometimes expressed as the mean fire return interval (FRI or MFRI). See point fire frequency, site fire frequency and area fire frequency. Fire frequency class – the range of fire frequency in a particular fire compartment. For mapping fire frequency it is useful to partition the fire frequency gradient into classes. Those I have found useful approximate a 2x geometric progression: 1-3 year fire-return interval, 4-6 years, 7-12 years, 13-25, 26-50, 51-100, 101 to >300 and nonpyrophytic. A fire frequency map often requires having a “Variable” fire frequency class, as for beaver wetlands, and other intervals may be appropriate depending upon the landscape. Fire frequency indicator community – a plant community limited to a specific fire frequency regime (such as canebrake, which reaches its best development with fire-return intervals of 2-8 years). Such communities can be used to map original landscape fire regime where remnants persist or there are historic records such as those on early survey maps. Fire frequency indicator species - plant species limited to a specific fire regime. Venus flytrap, the most fire-dependent species known, requires a mean fire return interval of at least 4 years and so is an indicator species for a 1-4 years fire frequency. Such species can be used to map original landscape fire regimes where the indicator species remains or where herbarium or other historic records exist. Fire frequency relations of few species have been documented and this is one of the major gaps in knowledge needed to manage landscapes for survival of the full diversity of species they contain. Fire frequency gradient - change in fire frequency across a fire-tension zone, where a gradient of topography, soil texture, or soil moisture, reduces intensity of fire as the fire moves along the gradient. Fires originating from the frequent-fire end of the gradient will penetrate different distances into the gradient toward its fire-sheltered end before going out (see fire effects gradient, which subsumes effects of both frequency and intensity). See also relative fire frequency. Look for the fire frequency gradients in any landscape. Fire frequency map – a map in which past fire frequency can be assigned to each fire compartment, taking into account historical records of presettlement tree species, vegetation types or fire occurrence, remnant fire-frequency indicator species, fire frequency indicator communities, natural firebreaks, fire filters and pathways for fire flow in the landscape. Fire inhibitor species - species with characteristics of litter, plant growth form, plant structure or chemistry which inhibit spread of fire or fire intensity. 3

Fire-deferred succession - the change, after removal of fire, from communities with only one or two vegetation layers to multistoried woody vegetation with few herbs. Fire legacy – pattern of scattered pyrophytic trees and fire dependent herbs persisting in a fire suppressed landscape that may now be composed largely of species different from those present under the original fire regime. Fire path - a continuous conduit of grass or other fuel, unbroken by rock outcrops, barren areas or other firebreaks or fire filters. The longer the fire path the higher the fire frequency. A longer fire path increases fire frequency in three ways: 1) it increases the size of the fire compartment, providing more area to receive lightning ignitions and 2) small fires can join in from either side of the path. Fires on either side may have discontinuous fuel, less flammable fuel or topography not conducive to fire spread and so might tend to be small but once they touch off the fuel in the fire path it can take off and run. 3) a fire path may also extend toward an ignition source such as an isolated mountain where orographic uplift generates lightning ignitions on the downwind. The path may collect those fires, conducting them downwind and augmenting the fire frequency throughout the length of the path and other downwind fire compartments it contacts (a fire [path is also a fire compartment). Fire paths can be temporarily interrupted as in the historical landscape when a herd of bison, attracted to the more lush grass along a fire path in shortgrass prairie grazed it down and moved on. Fire reach – The length of continuous fuel between sharp bends in a narrow fire corridor. The longer either the reach, the higher the fire frequency. Fire-refugial species - species such as beech, magnolia, cottonwoods, silver fir and a great many others which, in frequent-fire landscapes, are confined to naturally fire sheltered or fire-free sites. Fire regime types – may result from the classification of fire regimes in terms of frequency, severity, season of burn, periodicity (nonrandom or predictable, irregular, polycyclic), and ecological fire effects. Fire resistant community - community with collective characteristics which offer little fuel for spread or intensity of fire (high canopy, lack of shrub and herb layers, rapidly flattening and decomposing litter. Fire resistant species - species like most pines and oaks with adaptations like thick bark to resist damage and rapid self-pruning to minimize ladder fuels. Fire-return interval (or just fire interval) – time between two fires on a specific site or in a specific community type. Usually expressed as the mean fire interval (MFI or MFRI). Fire season – the time of year, in an undisturbed natural landscape, when most fires occur. It may be focused on a narrow window as in the six-week, mid July to end of August lightning ignition season in the shortgrass prairie of eastern Montana or the seven month, March through September season with a peak March and April in the southeastern U.S. And historically in this region the fire season may have been bimodal with a second peak caused by Native American burning in October and November. Fire season is characterized by two main components, the months in which the most ignitions occur and those in which the greatest area burns. They are related but the peaks for each are usually offset by a few weeks. Fire severity – one silvicultural interpretation applies the term by the degree of effects on vegetation as in High Severity (most stems killed to the ground, Low severity (most stems survive) and Mixed Severity (intermediate effects) . Severity is not necessarily related to intensity. A low intensity fire may pass quickly through a stand, generating only a small amount of heat but may produce severe fire effects as when duff smoldering kills roots or girdles the stem leading to loss of canopy trees. Fire shadow - area on the downwind side of a body of water or other natural firebreak, often sheltering 4

fire refugial vegetation in an otherwise frequent-fire landscape. Fire sheltered – adjective applying to a site in a pyrophytic landscape where topography, moisture of or other factors may reduce fire frequency or fire intensity below that of the surrounding landscape. One example would be an upland on the downwind side of a ravine containing wetland vegetation that slowed or sometimes topped the passage of fire. Fire spread - the rate of fire flow (customarily expressed in chains per hour). Firestream – a dominant pathway for fire flow in a particular landscape. Multiple firestreams may converge to create a hot spot of unusually high fire frequency. Fire tension zone - the intermediate region between fire-exposed and naturally fire-protected parts of the landscape. Local fire-tension zones may separate disparate vegetation types only a few meters apart. In regional fire tension zones, fire frequency may decline gradually over several kilometers in a flat landscape where fire can only approach from one direction. Fireline intensity – amount of heat released at head of fire, expressed in BTUs/ft/sec. Fireshed – region around a natural area having vegetation that may lead fires into the site. Fixed lightning generator (orographic lightning generator) – a tall landscape feature such as the Mogollon Rim in Arizona or the Black Mountains in North Carolina that lie transversely to the prevailing winds and have a long unobstructed run upwind so that when moisture conditions are right, they provide uplift to generate thunderheads that carry lightning into the downwind landscape, often with little associated rain. These create mappable and geologically permanent situations for frequent fire downwind and may sustain habitat for rare species dependent upon frequent fire for millions of years. Lightning generators include also single isolated mountains or small ranges that act as points for generating lightning, starting fires within a zone typically less than 100 miles downwind. Flanking fire – fire expanding sideways or perpendicular to the wind direction and main flow of fire. Flanking fires typically move slower than headfire and faster than backing fire and so have intermediate effects. Forest - community with a closed or nearly closed tree canopy. Usually applies to naturally fire-protected communities, or communities that experience only light surface fires, or vegetation like jack pine, sand pine and white cedar, that experience catastrophic fires, but with a fire-return interval of 30-300 years. Flame length – length of flames at head of fire. Can vary from a few inches in a grass fire to 40 feet in canebrake, to 1,000 feet in an upslope crown fire in western conifer forest. Frequency-dependent species or communities – plants that require fire at a particular frequency for survival or full expression. Examples are Venus flytrap, which requires fire at 1-3 year interval for survival, and canebrake which requires fire at 2-8 years intervals for maximum stem density. See fire frequency indicator species. Fuel equilibrium – Time after last fire by which fuel decomposition equals fuel accumulation. Fuel models – Fuels characterized by the fire carrier type, such as short grass, tall grass, forest litter, or chaparral, and the expected fire behavior in each. In one system, the National Wildfire Coordinating Group currently recognizes 12 fuel models. Fuel species – the principal species in a community which contribute to fire spread. Fuel structure – arrangement of fuels in relation to flammability. The spatial arrangement affects fire intensity: fuels that lie compactly will be less flammable and burn with lower intensity than cross-stacked twigs and limbs or light airy fuels such as pine needles draped over small branches on saplings. Fusain - fossil plant fragments charcoalized by fire at time of preservation. A component of coal and sedimentary rocks characterized by black color, silky luster and fibrous texture. Fusain is 5

considered evidence of natural wildfires in paleoforests (Calder et al. 1993, Jones and Rowe 1999). Grass reduction fires – surface fires in light, airy fuel, predominantly grasses. Intensity varies with density and length of grass, as well as temperature, relative humidity and wind. Ground cover – the fuel matrix of graminoids and forbs on the surface. Ground fire – fires in organic substrates such as peat (as opposed to surface fire in litter and other fine fuel). Fire may burn down to the water table or far enough to expose mineral soil, or may burn deeper than the seasonal high water table during dry spells, initiating ponding when rains return. Head fire – the portion of the fir perimeter moving rapidly downwind. Historical range of variation (HRV) – for fire frequency, the range of likely fire intervals for a particular fire compartment (site). For example a site with a mean fire frequency of four years might have an HRV of 2 to 8 years, with the range being an estimate of the interval in which 90% of fires occur. Ignition fans (orogenic lighting ignition fans) – the area downwind from a fixed lightning generator within which lightning strikes may produce fires. These areas are usually fan shaped, expanding downwind from the source (usually a single mountain, a range or an escarpment that interrupts the flow of moisture laden air) to a distance of about 100 miles, beyond which the temporary orogenic storms collapse below their ability to stop produce lightning. Ignition saturation – A condition where ignitions are so frequent that fuel is ignited almost as soon as enough accumulates to carry fire. Some areas, such as southeastern savannas in the largest fire compartments, and parts of the Southwest, appear to have experienced ignition saturation from lightning alone. After invasion of the New World by man some 12,000 years ago, new areas such as the Willamette Valley of Oregon became saturated as the result of annual ignitions by humans. Ignition source – most fires in the presettlement landscape were initiated by lightning or Native Americans. Rare instances of spontaneous ignition (in dead marsh grasses) have been reported, as well as fires ignited by volcanoes and sparks from falling rocks. All Native American ignition was superimposed on the natural lightning ignition background. Intensity – Fire intensity is a measure of the rate of release of heat. It includes both radiant and convectional heat and is an important contributor to fire severity. Ladder fuels – fuels such as vines, dead twigs, dead branches and dead pine needles draped on understory hardwood limbs that may carry surface fire upward into the canopy. Landscape fire ecology - the study of physical and biological factors that control the frequency of fire, the movement of fire in a landscape, and the effects of fire on vegetation and other organisms. Fire at any point in a landscape is a function of topography, other landscape factors (such as fire compartment size), vegetation, climate, animals and time since last fire (Fire = f(R*T*C*A*V*L)). Land surface form – a classification of topographic features according to 1) the percent of the landscape that is flat or only gently sloping, 2) amount of local relief from the stream bottoms to the ridge tops, and 3) whether the flat or only gently sloping parts are located on uplands or in bottoms (Hammond (1964). This gives a way to draw boundaries between parts of the landscape that have similarities in fire compartment size, density of fire filters, and density of firebreaks. This in turn permits mapping generalized fire regimes of major geographic regions (Frost 1998). Light environment – fire effects on species diversity in the understory and herb layers are largely due to its effect on the light environment, by opening up or maintaining access to sunlight. Where there is sufficient precipitation more open landscapes succeed to multistoried woody vegetation and litter buildup, eliminating the herb layer. The majority of fire dependent species occur in this 6

layer and their reliance upon fire to remove litter and duff and to provide light appear to be the primary components of fire dependence. Lightning fire season – the period of weeks when the greatest area of land is burned as a result of ignitions by lightning. Fuel continuity, seasonal fuel availability and moisture conditions related to drought are more important than lightning strike density. In southern Mississippi the lightning fire season peaks in May when dry lightning strikes are most common but all lightning strike frequency peaks in July (JS Brewer, in press). It is only necessary that there be some lightning ignitions when fuel conditions are optimum for fire spread. Lightning target – a topographic feature of the landscape that receives the highest lightning strike density in a particular landscape. May be an isolated peak or ridge, a prominent scarp lying perpendicularly to the prevailing wind during fire season, or an area of landscape just downwind from a topographic feature that creates orographic lifting and cumulus formation to generate lighting as the clouds move away downwind. Can be also be applied to any fire compartment; a peninsula might be called a small lightning target. Maintenance fire regime – a program of prescribed fires needed to simulate the natural fire regime after a site has already been restored using restoration burns (see). Marsh - wetlands dominated by emergent herbaceous vegetation, either grasses or broad-leaved herbs. Most graminoid mashes were frequent fire types in the original landscape, with fires spreading in from the now fragmented and suppressed uplands. Only a few types, such as those dominated by forbs like Sagittaria and Peltandra will not burn. Mean Fire-Return Interval (MFRI) – the mean of the times between fires on a specific site or in a specific community type, averaged over many years to give the typical fire frequency regime. Fire-interrupted succession - the change, after removal of fire, from communities with only one or two vegetation layers to multistoried woody vegetation with few herbs. Minimum dynamic area – the area of a landscape required to contain all phases of succession, all plant species patch types, all plant and animal species diversity, and enough populations and metapopulations of each as required to maintain the species in the landscape, under the natural fire regimes associated with the landscape (whether maintained by natural or planned ignitions). Mixed pine savanna - bilayered community with 2 or more pine species over a savanna herb layer maintained by frequent fire. Examples include any combination of Pinus palustris, P. echinata, P. serotina and P. taeda, often with a minor component of oaks and hickories in the southeastern US, or pitch pine (Pinus rigida) and Table Mountain pine (Pinus pungens) in the Southern Appalachians). Nonpyrophytic communities – relatively fireproof vegetation such as tupelo swamps (Nyssa aquatica), isolated vegetation clumps above treeline, talus slopes, rock outcrops, lava flows in the pioneer stages of succession, and species of deserts, playas and salt flats. This category also includes some arid grasslands lacking sufficient fuel to carry fire. Obligate fire frequency – the fire frequency range upon which a plant species is dependent for its survival. Obligate pyrophyte - Plant species dependent upon fire for completion of its life cycle. Oligopyric communities – Sites that ordinarily do not burn because of wetness or lack of fuel continuity, but which may carry a fire under extraordinary conditions of wind or drought. Orographic lightning generator – see fixed lightning generator. Orographic lightning ignition fans – the zone, often extending downwind for around 100 miles from a mountain serving as a fixed lightning generator in which lighting ignitions create a higher fire frequency than in the surrounding landscape. Pine marsh - bilayered fire community type of estuarine regions, such as those in the Southeast 7

dominated by loblolly pine over a graminoid layer, commonly Carex hyalinolepis or Chasmanthium laxum, or pond pine over a variety on marsh species, commonly including sawgrass (Cladium jamaicense). Point fire frequency – the fire-return interval for a specific point in the landscape. Usually obtained from a point sample, that is the number of fire scars on a particular tree. The point frequency is the most important frequency for survival of an individual fire-dependent plant. In dendrochronology, point samples are nearly always substantial underestimates of site fire frequency (see) because it is unlikely that every fire will scar a particular tree. Pocosin - evergreen shrub bog. May occur on organic or wet mineral soils under a wide range of fire- return intervals. Historically, and as used by Algonquian Indians, the term meant any open, relatively treeless plant communities, especially those in wetlands. Prairie - community consisting of only an herb layer, maintained free of trees and most shrub cover by various combinations of climate, soil type, soil moisture and fire. The term carries an implication of extensive area but is sometimes applied to treeless areas as small as 1 hectare in the southeastern U.S.). Presettlement fire frequency – In the western hemisphere, the prevailing fire return intervals at time of European settlement. Presettlement fire frequency was a composite of fires resulting from ignitions by Native Americans, against a background of lightning ignitions, which varied greatly in different parts of the presettlement landscape. Portions of the original landscape can be identified where either ignition source was dominant. Presettlement vegetation – in the western hemisphere, the natural vegetation which existed, under natural fire regimes, at time of European settlement (ranging from 1565 in Florida to around 1890 in remote parts of the western U.S.). Prevailing winds during fire season – over hundreds of years, the dominant wind direction by which fire most often approaches a particular fire compartment. Probability of ignition – the probability that a fire will continue to burn if an ignition source, such as lightning, occurs. Calculated from fine fuel moisture, temperature and degree of shading. Pyrogenicity, pyrogenic - influence of vegetation on fire behavior, acting through factors like ignitability of dead and living vegetation, and physical and chemical characteristics of dead fuel produced by vegetation. Pyrographic - relating to environmental and biotic factors contributing to fire flow, fire frequency and fire intensity in the landscape. Pyrography - mapping fire frequency, fire behavior or fire vegetation in a landscape. Pyromorphology – topographic and other landscape features that affect fire frequency, intensity and fire effect by creating areas that are naturally fire sheltered or fire exposed. Pyromosaic - Shifting or stable mosaic of plant communities in which the dominant species in any one patch depends upon conditions at time of last fire. For example, in southeastern peatlands the same site may have had, in different decades or centuries, white cedar, pond pine, bay forest, swamp black gum, baldcypress, or pond cypress, depending on whether the level of the water table at time of fire caused the fire to remain on the surface, burn shallowly into the peat, destroying the seed bank, or burn deeply, pooling water. Pyrophilic – fire adapted or benefiting from fire. Pyrophoric – refers to a fuel type which tends toward spontaneous combustion. Some materials like coal have a temperature at which they will continue to heat spontaneously until combustion occurs (if oxygen is present). Pyrophyte or pyrophytic species - plant species dependent on fire for completion of their life cycles (obligate) or adapted to survival under natural fire regimes (facultative). 8

Pyrophytic landscape - landscape dominated by vegetation that owes its structure and species composition to fire. Pyrophytic species - plant species dependent on fire for completion of their life cycles (obligate) or adapted to survival under natural fire regimes (facultative). Pyrophytic woodland - bilayered community with the tree canopy and herb layer being the principal layers, the understory being kept largely clear of woody species by fire. Tree cover may be up to 75%. Trees may be mixtures of oaks, hickories (within their range, especially mockernut hickory (Carya alba [tomentosa]) and pines. Pyrotone – the fire-tension zone between flammable upland vegetation and less flammable wetland vegetation, often only a few meters wide. Rate of spread – a characteristic of fire behavior, traditionally expressed as fire spread in chains per hour (66 feet/chain). Reach – a downwind portion of a fire path between two bends. Reference Conditions – the spectrum of natural features of a landscape before European settlement, including all the factors of both vegetation and fire regimes. Reference Fire Regime – the historical frequency, severity, season of burn, and periodicity of fire for a particular site type or vegetation type. Reference site – a site used as a model for restoring natural vegetation and fire regimes to disturbed sites. Ideally a reference site has entirely natural vegetation which has experienced a natural fire regime continuously with only minor interruption since before European settlement. A reference site should be as near to the restoration site as possible, have the same soil type, slope, aspect and other biophysical characteristics, and occur in the same position in the fire landscape, that is, with the same degree of access by fire. Relative fire frequency – the fire relation of one point in a landscape to adjacent points. Even without knowing actual fire frequencies, a relative fire frequency map can be constructed for any landscape by identifying the most fire frequent and the most fire sheltered points in the landscape. Between these extremes, relative frequency for intermediate points can be interpolated and assigned to fire compartments. This is done by interpreting the influence of natural firebreaks, fire bottlenecks and fire filters and then artificially chopping the fire frequency gradient into seven or eight segments having decreasing fire frequency. Constructing a relative fire frequency map is the first step in reconstructing presettlement vegetation and fire regimes. Restoration fire regime – the set of prescribed fires needed to restore a site to natural conditions before instituting a pattern of maintenance burns. Restoration burns may be intentionally hotter or cooler or more frequent than natural or may conducted out of the normal fire season. For example a site with fire excluded for 40 years may have developed an unnatural duff layer in which the trees have invested a good percentage of their fine root mass. Burning such a site when the duff layer is dry can destroy enough root mass to kill normally fire resistant trees such as oaks and even longleaf pine. In western fire types such as ponderosa pine, decades of fire exclusion may lead to a buildup of species such as Rocky Mountain juniper (Juniperus scopulorum) and unpruned, low, dead branches. A series of three light, cool season burns might be used for progressive removal of shrubs or of the duff layer in the eastern case, allowing trees time to relocate their fine root mass into the mineral soil. After this is accomplished a regime of normal, warm season maintenance burns may be established. In the majority of cases, the prescription switches from the restoration phase to maintenance burns after a graminoid-forb layer becomes established (as in savannas, woodlands or marshes). In the case of canebrake, chamise, chapparal or pocosin, the maintenance phase may begin when excess tree cover has been removed and the stem density, height and species composition have been restored to a 9

condition within the historic range of variation. Savanna - a bilayered community maintained by fire, with an open tree canopy and an almost continuous herb layer. Tree cover may be up to 50%. Scale of fire – PointSite (= fire compartment)FireshedLandscapeRegion. Serotiny – production of cones that remain closed until stimulated by fire (or passage of time) with the result that seeds are disturbed on bare ground with plenty of light after the fire passes. Severity – see fire severity. Shrub reduction fires – intense fires in shrub-dominated communities that reduce all stems to the ground. Some shrubs such as chamise (Adenostoma fasciculatum) and big sagebrush (Artemisia tridentata) may be mostly killed outright and have to start over from seed but the majority are prolific resprouters. Site fire frequency – the mean fire return interval for a particular fire compartment. Usually derived from a fire frequency map or from a composite fire scar chronology using fire scar dates from a number of trees within the compartment. This is the most ecologically useful estimate of fire frequency since it is the actual rate of fire experienced by all vegetation, including the herb layer, in a particular fire compartment (compare point samples and area samples of fire frequency). In frequent fire communities, a fire may pass through a stand without scarring a single tree. In such communities fire scar chronologies would be expected to substantially underestimate fire frequency. In such communities, many species of the herb layer (where most of the species diversity is found) may require for survival, a higher fire frequency than that recorded by the trees. Stand-replacing fire – fires which kill trees to the ground. There are two types: crown fires and lethal understory fires. Understory fires may kill trees by bole heating or by destroying the roots. Some canopy tree species may be killed outright while others have the ability to resprout. Surface fire – light fires in hardwood leaf litter, conifer litter, light grass and forbs. Understory reduction fires – fires in multistoried stands, intense enough to clear out everything in the understory, including subcanopy trees, but leave most canopy trees intact. Understory thinning fires – fires in multistoried stands which thin the shrub and subcanopy layers without killing everything beneath the canopy. Within-compartment fire effects – variation in fire frequency within a single compartment may be predicted by comparing fire exposed points in the compartment with partially fire sheltered points. Fire frequency and intensity can be expected to be higher on dry south slopes and ridges, at points downwind from the prevailing fire season wind direction, and points in the vicinity of a connector or window into another fire compartment—an additional source of ignition. Fire frequency can be expected to be lower near the upwind border. Xerophytic vegetation – plant communities in topographic situations or substrates dryer than the average for the regional landscape. Such dry vegetation can burn more or less frequently that mesophytic vegetation. While fuels might be expected to be dry, on poor soils they may be too sparse to carry fire. In such cases they often serve as fire filters, reducing fire frequency downwind.