}`HPLJ2Ö}.MT 1.00" .MB 1.00" .ls2 .oj on ŠÜÀ´ð  úú! TREE-RING RECONSTRUCTION OF GIANT SEQUOIA FIRE REGIMES  êê! FINAL REPORT  ±±! INTRODUCTION  „„  This  project  formally began in June 1988 and ended in May 1991. An  earlier  feasibility study  was conducted in 1987 (Swetnam and Baisan 1988), and a current expanded fire  history and  fire climatology study funded by the National Park Service Global Change program  began in June 1991 (Swetnam 1991). All of these projects have focused on improving our understand ing  of  past  fire  regimes  in  Sierra Nevada forest ecosystems.  This  final  report  will  serve  to document  and  transfer  the  basic data to the National Park Service  and  to  summarize  major findings of analyses completed during the period from 1987 to 1991.  „„  The  main  objective  of this project was to reconstruct past fire history  in  different  giant sequoia  groves  distributed along a north-south transect from Yosemite  National  Park  south ward  to Mountain Home State Forest. We planned to develop millennia-length  fire-scar  chro nologies  for  different sequoia mixed-conifer groves and to compare the fire  histories  (fire  fre quency, size, intensity, and season) within and among these groves. Earlier fire-scar studies  in the  Redwood  Mountain Grove by Kilgore and Taylor (1979) demonstrated  that  pre-settlement (pre-1875)  surface fires were relatively frequent with two to three year mean fire  intervals  within watersheds  of about 800 to 1000 ha., and five to nine year mean fire intervals in smaller sites  of 3  to  16 ha. These fire-scar chronologies, based upon specimens from pines,  fir,  and  incense cedar, were used to estimate fire frequency variations within two watersheds back to about A.D. 1700, although a few samples had scars extending back into the late 1400s. Our goal has been to  spatially  expand and temporally extend the knowledge base of fire history to  other  sequoia mixed-conifer groves in the Sierra Nevada by sampling ancient fire-scarred sequoias.  „„  The impetus for this work was an identified need for a broad base of long-term fire histo ry  information  in  sequoia mixed-conifer forests (Christensen et al. 1987).  A  program  of  pre  ð63 Šscribed  burning had been conducted within the groves in Sequoia, Kings Canyon, and  Yosem ite National Parks since the late 1960s. Following a detailed review of the state-of-knowledge of sequoia  mixed-conifer  fire  ecology  and  the  fire  management  programs  in  Sequoia,   Kings Canyon and Yosemite National Parks the Christensen panel strongly endorsed the continuation of  prescribed  burning as necessary for reintroduction of fire as a natural process in  these  for ests.  However, the report stated that more information was needed on past fire  occurrence  in giant  sequoia  groves, including the range, variations, seasonality, and temporal  trends  in  fire regimes  in  different groves. The role of lightning versus Indians as a source of fire on  the  pre- settlement  landscape  was also mentioned as a priority research topic. This  baseline  informa tion  would  be valuable for better understanding of sequoia fire ecology,  forest  dynamics,  and the historical development of sequoia groves. Improved knowledge in these areas should  help develop better fire management plans by suggesting burning strategies and objectives to  more closely mimic the long-term historical fire process.  „„  This  information  will also serve an important role in educating and informing  the  public and  managers  about  the natural history of climate, fire, and people  in  sequoia  mixed-conifer forests. The amazing length and detail of the sequoia fire history record that we have  complied has  already  generated tremendous interest and enthusiasm among scientists  and  the  public. Our current plan is to revise this report and submit the major findings as one or more papers  to peer  reviewed  scientific  journals.  In  addition to the analyses  and  results  reported  here,  we expect to conduct other analyses using the sequoia tree-ring data base and to report all  results in peer reviewed scientific journals. Appendix E lists some of these analyses, planned publica tions, and existing papers and publications derived in part from this work.  %%! METHODS Study Areas and Sample Trees  „„  We  selected five sequoia groves for fire-scar sampling in Yosemite, Kings  Canyon  and Sequoia National Parks and in Mountain Home State Forest (Fig. 1 and Tables 1, 2).  ð6 2  Š.pn 6 ..Fig. 1, pg. 5 These  particular  groves  were  selected because they  were  accessible,  they  contained  dead sequoia stems or stumps that could be sampled, and they were important management  areas. The Giant Forest and Mountain Home groves are among the largest of about 75 extant  sequoia groves  (Fry and White 1938, Rundel 1972). We sampled 10 different groups of trees within  the Giant  Forest,  but here we will only report the results from the concentrated collection  in  Circle Meadow  (Fig. 2). The other groups of specimens from this grove (Log Meadow,  Huckleberry, Giant Forest Village, and Hazelwood) are still being analyzed. Similar clusters, or groups of fire- scarred  trees, were sampled in sub-areas of other groves (Figs. 3, 4, 5, 6, Appendix B  includes a  table  listing  group names and assigned database numbers). The  purpose  of  this  strategy was  to  provide  better  composite  estimates  of fire  occurrence  at  point  locations  within  the groves. Previous fire history studies found that individual trees often did not provide a complete estimate  of the total number of fires that occurred at the location of the tree, but  samples  from clusters  of trees in relatively small areas (within about 1 ha) improve the chances of  document ing  the  occurrence  of  smaller  fires (Kilgore and Taylor  1979,  Swetnam  and  Dieterich  1985, Swetnam et al. 1990, Baisan and Swetnam 1991).  „„  The  areas  of  sampled groups and groves listed in Table 1 were defined  by  a  polygon whose  vertices were the outermost sampled trees in a two-dimensional plane. This polygon  is termed  a  "convex  hull"  because  all  of the vertices are convex  from  the  center  of  the  area. Hence,  the  vertices and perimeter of the polygon were usually defined by fewer trees  than  the total sampled number of trees, with the trees not forming a convex vertex being included  within the  polygon.  The  Atwell  and  Mountain Home maps (Figs. 5  and  6)  show  examples  of  the convex  hull based on all sampled trees. This scheme was arbitrary in the sense  that  topogra phy  or  other  spatial  aspects of the terrain were not explicitly considered,  but  it  was  a  useful conservative  estimate  of  the  minimum  areas that the  fire  history  might  represent.  Using  a planimeter,  topographic  maps, and the National Park Service's sequoia  tree  inventory  maps (Western Tree Service 1970) each group area was measured separately, then the convex hull of the  total  area  based on all sampled trees within each grove was  measured.  ð6 2  Š.pn  10 .. Figs. 2-9, pgs. 7-9 The  area  estimates from Mariposa and Mountain Home may be somewhat less  accurate  than the other groves because detailed stem maps were not available for these groves.  „„  All  sequoia  fire-scar  samples  were taken only from dead trees.  In  Circle  Meadow  all dead trees were either snags or logs. Sequoia stumps and logs from late-19th century  harvest ing  were  sampled in the Big Stump, Atwell and Mountain Home groves. Two of  the  Mountain Home  stumps  were  from  logging  in the 1930s and 1950s,  and  two  of  the  Mariposa  Grove specimens  were  from  stumps near the site of an old lodge; the remainder  were  from  logs  or snags.  „„  The  majority of the remnant stumps were missing most of the sapwood which  decayed much faster than heartwood. We estimate the sapwood usually included the outer 10 to 20  cm of  growth  and the last 100 to 200 growth rings. Despite loss of the outermost fire  history  con tained  in the sapwood in most specimens, we were able to obtain some samples covering  this period  (1800s  to  mid  or late 1900s) in all groves from portions of  stumps  that  still  contained sapwood, or from more recently dead snags, logs, and stumps.  „„  Partial  cross  sections were obtained with chainsaws. We typically used bar  lengths  of about  1.4 m, but for some cuts we used a 2.1 m bar. Old giant sequoias often contain  numer ous  fire  scar  cavities around the stem (Fig. 7). We searched for trees that  had  multiple,  deep cavities  with  well preserved fire-scar ridges visible along the inner walls of  the  cavities.  Some trees  had a characteristic fire-scar type that we termed "eye-socket" fire scars (Fig.  8).  Parallel horizontal  cuts  approximately  5  to  10 cm apart were made through  the  wall  of  the  fire-scar cavity  or  eye-socket,  then vertical cuts were made to free the partial section.  The  size  of  the sections varied greatly, but the largest pieces were about 1 x 2 m in width and length. Several partial  sections were taken per cavity or eye-socket, and up to 14 sections were taken  from  all sampled  cavities  on each tree (see Table 2 for list of total number of trees and  partial  sections sampled in each grove).  „„  In  addition to the large partial sections through the fire-scarred boles, we  also  obtained radial "v-cuts" or "plunge cuts" from all fire scar sample trees. These samples were very useful  ð6 2  Š..Figs. 7,8, pgs.11-12 .pn 13 in  crossdating  the  fire-scarred  sections taken at lower levels on the  bole  because  they  were relatively  free of ring distortions caused by fire scars. The v-cuts, originally described  by  Dou glass  (1919),  were radial sections taken from the tops of cut stumps. They  were  obtained  by making two intersecting cuts approximately in the shape of a "V" along a radial axis from pith  to outermost  rings  on  the  top  of a cut stump. A radial plunge-cut was  a  long  and  narrow  pie- shaped  section  taken from a log or standing snag by "plunging" a roller-tip  chainsaw  into  the bole  to  create four intersecting cuts that met at maximum depth in the bole.  A  skilled  sawyer (Chris  Baisan)  was  usually  able to obtain plunge-cuts of 10 to 20 cm thickness  and  1  to  1.5 meters in length, and in a couple of instances, over 2 meters length.  „„  During  collection  a  detailed specimen card was filled out for each  sampled  tree.  This card included a sketch drawing of the tree and relative locations of all partial sections, NPS  tree inventory numbers, and miscellaneous observations of the tree and site. Each fire-scar  sample tree was tagged with an aluminum or brass tag with our identification numbers.  „„  The  sampling  of  the different groves required 3 to 4 weeks of work in each  of  the  four summers from 1987 to 1990. All sections were hauled to nearest roads and then transported  in large  rental  trucks  to our laboratory in Tucson. We estimate the total quantity  of  the  sequoia sections  transported  exceeded  100 cubic meters in volume, and more  than  4,500  kilograms (dried)  in  weight.  This  is  probably the largest single  collection  of  tree-ring  specimens  ever obtained  and  analyzed. The bulk of this collection is currently in storage at  the  National  Park Service's  Western  Archaeological Center (WAC) in Tucson, Arizona. A complete  inventory  of specimens  at WAC is on file there and at the Tree-Ring Lab in Tucson. This list, and  copies  of original  field specimen cards for each sampled tree have been delivered to Sequoia  and  Kings Canyon National Parks. Laboratory Analyses  „„  Some  specimens  were composed of many fragments that had to  be  pieced  together, glued,  and  mounted  on plywood sheets. The sections were sanded  with  belt  sanders  using  ð6 2 Šsuccessively  finer  grits  (up to 400 grit). The specimens were then  examined  on  large  tables with  variable  power  binocular  microscopes (10X to 30X) mounted  on  long  extendible  arms. Lighted  jeweler's  magnifying  glasses  mounted on extendible arms  were  also  useful  for  the dating work.  „„  The  tree  rings  on the sections were crossdated using a  combination  of  skeleton  plot procedures  (Stokes  and Smiley 1968, Swetnam et al. 1985) and ring-width  pattern  memoriza tion. The latter approach, originally described by A. E. Douglass, is based on the memorization of  signature years (i.e., smallest ring years) that are consistent within the trees and region  stud ied.  Douglass  (1919) originally developed a 3,221-year tree-ring chronology  from  v-cut  radial sections  from  giant sequoias that had been felled in logging operations. We  crossdated  the first specimens we collected with the Douglass ring-width chronology, and later we used sever al  new  chronologies  developed  from v-cuts and plunge-cuts  obtained  from  different  groves (Hughes and Brown 1992, Brown et al. in review). The radial sections from individual trees were often dated first and then used as a key to the dating of the fire-scarred partial sections from the same tree.  „„  Fire  scars  were  dated by observing their position within  the  crossdated  annual  rings. Calendrical dates were noted for all fire scars appearing on each partial cross section.  Appen dix  A is the Data Input Documentation that served as a guide for identifying and entering in  our data  base  all  of  the  observations of dated fire scars.  We  found  that  several  other  tree-ring indicators provided additional important evidence on timing of past fires because these features were    very   consistently   associated   with   actual   fire   scars.   These   indicators   -    growth releases/suppressions, resin ducts, and expanded latewood - will be described and  discussed in  the  results section. We refer to them as "other indicators". Combined evidence  of  past  fire dates,   including   fire   scars  and  the  other  indicators,  are  referred  to   collectively   as "fire indicators".  „„  In addition to calendrical dates of fire scars we also noted the relative position of the  fire scars  within the annual rings so that we might infer seasonal timing of past fires  (Dieterich  and  ð6 2 ŠSwetnam 1984, Baisan and Swetnam 1990). Appendix A includes a description and  illustration of  the  classification  of intra-annual fire scar position. This identification  was  based  on  visual observation  of  the position of the fire scar - a fire-caused lesion - on the annual  ring  boundary (dormant  season scar), within the earlywood-type tracheid cells (earlywood scar), or within  the latewood-type tracheid cells (latewood scar). In most cases it was possible to further  subdivide the  earlywood into approximate thirds and subclassify the scar according to its  position  within each third (early-earlywood, middle-earlywood, late-earlywood).  „„  On  any  given tree a specific fire date was often recorded by numerous fire scars  in  the same  ring on different partial sections taken from different locations on the bole. Hence,  many observations  of  scar  position  were usually available for each fire  date  on  each  tree.  Dating sheets  containing  all  observations  were then summarized for each  tree  on  a  master  dating sheet.  The  data  were  then entered into a database (dBase IV)  via  a  customized  data  entry program  developed  by  A.  C.  Caprio. Appendix A describes the  data  fields  included  in  this database.  „„  The  consensus, or most commonly observed intra-annual position for each fire date  on each  tree  was entered in the data base. Usually, the majority or all of the  observations  of  the same scar date were classified as the same relative position. If a different position for the  same fire  date was observed on other partial sections from the same tree it was noted as  an  alterna tive position in the data base.  „„  Following  data  entry  a  check program was employed to screen  all  data  for  possible errors,  such  as duplicate or incorrect dates. All consecutive-year fire dates  recorded  within  a grove,  and  the specimens recording them, were identified by this program. All of  these  cases were rechecked because the single year differences could have been due to dating errors,  data entry  errors,  or  problems  relating  to  observation of  the  fire  indicators.  In  these  cases  the specimens  were rechecked by the dating team, and the identified dates were confirmed,  modi fied,  or  deleted if the initial observations were deemed to be unreliable. Altogether,  the  dating and checking procedures encompassed a tremendous amount of laboratory working hours.  ð6 2  ŠData Compilation and Analysis  „„  Fire  frequencies  (number of fires per time period), mean fire intervals (MFI,  or  average number  of  years  between  fires),  and  ranges and  standard  deviations  of  fire  intervals  were computed  for  different  periods  for each grove. Century-long  periods  were  typically  used  in summary computations. A common period of A.D. 500 to 1900 was used in most of the  inter- grove  comparisons  because  this period was generally well represented by the  samples  in  all groves.  „„  Changes  in  fire  frequency through time were explored by computing  and  plotting  fire frequency during moving periods. The moving periods were overlapping time periods of  differ ent  length  (50  and  20 years) during which the total number of  recorded  fires  were  summed. Each  sequential  value  was  the summation of fire events in the  time  period  lagged  one  year forward from the previous period. The fire frequencies in these moving periods were plotted  on the  approximate  central-year  of the period. For example, the 50-year moving  period  fire  fre quency  value  plotted  at A.D. 1600 was the total number of fires recorded from  1576  to  1625, and the value plotted at 1601 was the total for the period 1577 to 1626, and so on.  „„  MFIs  and fire frequencies were computed for (1) fire events recorded by fire  scars,  and (2)  fire events recorded by all fire indicators (fire scars and other indicators). Additionally,  MFIs were  computed  for fires recorded by two or more trees. Interpretations of  these  different  MFI estimates will be discussed.  „„  Observed temporal changes in fire frequency may be related to changes in sample  size (number of fire scarred trees) or size of the study area (Arno and Peterson 1983). Individual  fire scarred  trees  probably did not record all of the surface fires that occurred in  their  vicinity,  and some  small  fires may have burned only around one or a few trees. Thus,  larger  sample  sizes increase  the probability of documenting such fires. Fire scar collections distributed over  larger study areas also have a greater probability of recording other fires that burned different portions of the groves.  ð6 2  Š  „„  We  estimated the effect of decreased sample sizes (and indirectly, sample area) on  fire frequency  by  a bootstrap method. First we computed the fire frequency during a  base  period (A.D.  1000  to  1599)  that included a relatively constant and maximum  sample  size  in  all  five groves. Next, we removed randomly selected tree records from the sample set and  recomput ed  the  fire  frequencies.  Each  removal  of  "n" randomly selected trees  (from  n  =  1  to  n  = maximum  sample  size - 1) was repeated 50 times. The mean fire frequency  and  95%  confi dence  intervals  were  computed  for the 50 repetitions for each  reduced  sample  size  in  each grove.  Plots of the mean fire frequency estimates and confidence intervals for reduced  sample sizes  were  produced for each grove. These estimates were compared  with  actual  centennial fire  frequencies and the corresponding sample sizes. This provided a means for determining  if centennial  patterns  of  fire  frequency changes were higher or lower than  the  base  period  fire frequency, while also accounting for reduced sample sizes.  „„  Relative  positions of fire scars within annual rings were summarized for each grove  and for  all groves combined. Changes in fire scar position through time were also  summarized  by computing and plotting percentage occurrence in each classification by century-length periods. Comparison  of  scar position data and recent observations of cambial  phenology  of  sequoias (Parsons,  pers.  comm.) allowed us to make some preliminary  inferences  regarding  probable seasons of past fire occurrence.  „„  Synchrony  of fire dates among the five groves was analyzed to determine  if  large-scale (regional) climate patterns were important to fire occurrence. The null hypothesis stated that no significant  co-occurrence  of fire events was observed among the five different  chronologies  of fire dates. Since the groves were widely separated in space, with little possibility of fires spread ing between them, any statistically significant co-occurrence of fire dates would suggest that  an exogenous factor, or factors, were important in synchronizing the fire occurrences. This  factor would  most likely be climate, as it influences fuel production, fuel moisture, and ignitions  (e.g., lightning)  on  regional  scales. Fire-climate relationships will be explored in  more  detail  in  the currently funded NPS Global Change project (Swetnam 1991a). ð6 2 Š  „„  The  significance of fire synchrony for different time periods in the groves was tested  via contingency  analyses  and  the  chi-square statistic (Gibbons  1976).  The  procedure  involved estimating  the  expected  number of fires that could occur by chance in all  five  groves,  in  four groves,  three groves, etc. during century-length periods, and then counting the  actual  number of  (observed)  fires  that occurred during these periods. The expected  joint  probabilities  were computed  from the observed fire frequencies in the five groves during each century.  The  con tingent frequency of no fires occurring in any grove was also tested.  ÀÀ! RESULTS AND DISCUSSION Long and Detailed Fire Records  „„  Crossdating  of the sequoia tree-ring specimens was highly successful. With only a  few exceptions,  all  of  the  partial sections and v-cuts crossdated  with  Douglass's  tree-ring  index chronology,  or  one  of  the sequoia chronologies we  developed  from  collections  in  previous years  during  the  course  of  this project (Hughes and Brown  1991,  Brown  et  al.,  submitted). There  are  large  numbers  of  well-preserved  stumps in  cut-over  groves,  and  there  are  also tremendous quantities of logs and snags in the groves. The accumulation of snags and logs  is partly  due  to fire suppression in this century; much of this material would probably  have  been consumed if periodic burning had continued. Our results show that most of this dead  material is dendrochronologically datable, and therefore environmental information contained in the tree rings is accessible and of high temporal resolution. In addition to detailed fire histories that may be  longer  then  those reported here, this material may also contain a variety of  other  types  of paleoclimatic and paleoecologic information that could be tapped in future scientific studies.  „„  The  dated  partial  sections we obtained from sequoia tress  contained  amazingly  long and detailed records of past fires. Individual partial sections from fire-scarred cavities often  had dozens  of  visible  fire scars. Over the past three millennia of the sequoia  record  we  observed more  than 3,200 fire scar dates and an additional 2,400 fire dates based on other indicators  on 544  partial  sections  from  86 trees (Tables 2 and 3). The numbers  of  observations  of  all  fire  ð6 2 Š..Tab. 3, pgs. 19-21 .pn 22 indicators  (scars  and  other) were actually many times greater  because  these  numbers  were compiled  by  tree  and each fire date was usually observed many times on  the  different  partial sections  from each tree. The average number of dated fire events per sampled  partial  section was 10.6, and the average number per tree was 63.8. Five different trees (BSU5, CMC3, MHF7, MHF2,  and GFV1) recorded more than 120 fire dates (fire scars and other indicators).  Table  2 includes a list of innermost and outermost ring dates for all sampled trees. All dated fire  events are  listed  by  year and grove in Appendix B. The oldest dated fire scar at  1125  B.C.  was  ob served on a section from the oldest tree we sampled (CMC3), which had an innermost ring date of  1238 B.C. This is an enormous snag near Cattle Cabin in the Giant Forest, and it is  now  the second oldest (dead) dated sequoia. The oldest dated stump is in the Converse Basin north  of Sequoia National Park; it has an inner ring date of 1306 B.C (Huntington 1914, Douglass  1919). (Note that the B.C. dates listed above are calendrical dates, while the "minus" dates listed in  the Tables and Appendices are tree-ring dates. Our tree-ring dates include a "0" year, while there is no  such  calendrical  year.  Hence,  specific  "minus" tree-ring  dates  listed  in  the  Tables  and Appendices are one year earlier than the corresponding calendrical dates.)  „„  In  addition  to  the partial sections we dated one full cross section from  a  large  mature giant  sequoia from the Giant Forest Village area that was cut in 1951 (Fig. 9). The pith  date  of this  7.3  m diameter section was 257 B.C. Eighty-three different fire scar  dates  were  observed and a total of 127 fire events were recorded by all fire indicators on this section. (Data from  this tree,  no.  GFV1 in our database, is not included in the Circle Meadow chronology  described  in this  report. These data will be included in a later paper that will be a synthesis of all  fire  history reconstructions  in the Giant Forest area.) The section was taken from a position less than  1  m above  the  ground  level.  In our sampling with chainsaws  we  often  observed  that  maximum numbers  of  well preserved scars appeared on sections taken within a half meter or less  of  the current  ground level. The display section also allowed us to date all visible scars in  all  cavities as  well  as  the most deeply buried scars near the pith. Our sampling  of  other  sequoia  stems was limited to the maximum length of the chainsaw bar. Where the fire-scar cavities were  deep  ð6 2 Š..Fig9, pg. 23 .pn 24 enough  we  were  also  able to sample near the pith of the trees, and  thereby  obtain  older  fire scar records.  „„  Several  remarkable specimens were obtained from the Big Stump Grove, with one  par tial  section  measuring about 1 by 0.3 m in length and width containing more than  60  fire  scar dates.  This section is currently on display in the Grant Grove Visitors Center. This tree  (BSC1, see Fig. 8) and another one (BSE2) each had a total of more than 80 fire scar dates on all partial sections,  and a third tree (BSU5) recorded a total of 153 fire dates (all fire indicators included)  - the maximum recorded by any tree. The lower boles of old sequoias often have large buttress es,  deep  fire scar cavities, and "eyesocket" fire scars (Figs. 7 and 8). We believe that  many  of these  fire scars originally formed at the apex of protrusions or "flutes" that form at the  bases  of sequoias.  The  apex  of  these flutes, which are areas of rapid ring  growth,  typically  have  thin bark  and  are  apparently  easily injured by fire. We repeatedly observed series  of  a  dozen  or more  completely  grown  over  fire  scars that had originally formed at the  apex  of  a  flute.  By completely excising eyesocket fire scars from several trees, and then sectioning these  samples in the laboratory, we learned that, in some cases the trees eventually cannot heal over the scars entirely,  and then these wounds acquire the characteristic eyesocket shape. The  buried  scars and eyesocket, a testament to the ability of sequoia to heal over and around wounds,  account ed in large measure for the very detailed and long fire record preserved by these trees.  „„  The spatial and temporal pattern of surface fires recorded by fire scars and other  indica tors  is  exemplified by a portion of the Circle Meadow fire scar chronology (Fig. 10).  We  judge that  the  number of partial sections and sampled trees with dated fire evidence at  each  year  in the  time  series  (sample depth) was generally sufficient to infer relative  changes  in  fire  occur rence  back  to at ca. A.D. 500 in all of the groves (Fig. 11). The Circle  Meadow  and  Mountain Home chronologies had sufficient sample depths for comparison back to about 1 B.C.  Sample depth,  and hence confidence in fire occurrence estimates, is lower during the period from  A.D. 500  to 900 in the Mariposa Grove than in the other groves (Fig. 11). Sample depths  tended  to decrease after the mid to late 1800s in all sites because we sampled only dead trees and  many  ð6 2 Š..Figs. 10,11, pgs. 25-26 .pn 27 of  these  were  stumps  of trees felled during this  period.  Nevertheless,  sample  depths  were generally  adequate  to  infer  relative  fire occurrence changes in the  groves  up  to  about  A.D. 1900.  Mountain Home and Big Stump, the two groves where stumps were  primarily  sampled, had  only  2 and 1 sample trees, respectively, extending into the twentieth century. Each  of  the other  three  groves had at least 5 trees extending into the twentieth century. All  groves  had  at least  six trees extending into the 1880s. Results of a quantitative test of the effects  of  changes in sample depth on fire frequency estimates is presented in a later section. Fire Scars and Other Fire Indicators  „„  Where  fire  scars  were present on a partial section one or more of  the  other  indicators (growth releases, resin ducts, expanded latewood) were almost always observed directly asso ciated  with  the fire scars (Fig. 12). The growth releases and resin ducts were  not  restricted  to the  area  immediately  around  a  fire scar, but could usually  be  observed  at  distant  positions around  the circumference of the stem. Fire scars were considered the "best evidence"  of  past fires,  and  so fire dates recorded by both fire scars and other indicators on the same  tree  were entered in the data base as fire scar dates for that tree. We frequently observed other indicators for  a particular date on partial sections without associated fire scars, and subsequently found  a fire scar at that date on another partial section from the same tree, or from another tree.  „„  There  are  several possible mechanisms for the association we  observed  between  the "other  indicators"  and fire events. We hypothesize that the growth releases  (increased  growth following  fires)  were  due  to  negative effects of the fires  on  competing  understory  plants  or changes  in soil properties (e.g., nutrients, moisture infiltration, temperature,  etc.).  Hartesveldt (1964)  also  observed these fire-related growth releases in a study of ring-widths  on  increment cores  taken  from  100  trees in the Mariposa Grove. He attributed  the  stimulated  ring  growth following fires to temporary reduced competition from damaged or killed "associates",  resulting in  improved  soil  moisture  conditions  for the more fire  resistant  sequoias.  We  are  currently  ð6 2  Š.pn 29 .. Fig. 12, pg. 28 studying  some  aspects of this phenomenon in prescribed burn areas within  Sequoia  National Park (Mutch and Swetnam 1991) and in Mountain Home State Forest (Swetnam 1991b).  „„  Traumatic  resin  ducts  are known to be related to stress or injuries  to  the  cambium  or phloem of conifers (Kramer and Kozlowski 1979). Hence, we suspect that this indicator  reflect ed  a direct injury to some part of the tree. Traumatic resin ducts appeared to be the  least  reli able  "other  indicator"  of  fire  because they were less  consistently  associated  with  actual  fire scars  then  the other indicators. Thus, a fire date was considered "questionable" if  it  was  only documented by resin ducts. Questionable dates were not included in the final compilation  and analysis  of fire occurrence reported here. Expanded latewood was characterized by  formation of  a  false  ring following the latewood. This feature was often observed  following  latewood  or dormant season fire scars. We hypothesize that this was either some sort of internal  physiolog ical response of trees to fire scarring, or it may be caused by a change in the environment of the tree  immediately following fires that stimulated cambial growth late in the growing season.  For example, an environmental stimulus could be fire-blackened ground in the late summer or early fall that resulted in soil warming followed by additional growth in trees that ordinarily would have entered cambial dormancy.  „„  An average of 35% of the total number of recorded fire dates within the five groves  were based on other indicators. Thus, the fire dates recorded by fire scars were, on the average, just under  two  thirds  of  the total amount of fire evidence. We believe that  most  of  the  additional dates  based  on other indicators were fires that burned in the vicinity of the recording  trees  but did not leave a scar, or we did not sample the scar that was inflicted upon the tree. The latter  is a  strong  possibility because of the difficulty in sampling all of the numerous fire  scars  that  are deeply embedded in the enormous boles of fire-scarred giant sequoias.  „„  It is possible, however, that a small proportion of the additional fire dates based only  on other  indicators  were  actually  caused by non-fire disturbances. We  did  observe  a  few  "fire" dates based on other indicators during the twentieth century with no known historical  reference to  fires  during  these  dates  in government  documents  or  other  documentary  sources  (i.e.,  ð6 2 Šnewspapers, etc.). Some of these dates may represent human-related disturbances other  than fire.  For example, we suspect that four early twentieth century dates based on other  indicators in  Mariposa  Grove  may  relate  to road building or construction  of  a  tourist  lodge  within  the grove.  Mechanical  injuries to the sampled trees (causing resin ducts), or felling  of  understory trees  (causing growth releases because of reduced competition) could explain the  occurrence of  these other indicators. Hartesveldt (1964) apparently also detected these  twentieth  century releases  and he attributed them to "vista clearing and removal of shrubs" by park managers.  It is  also possible that some of these dates were fires that simply were not documented  by  early administrators of the parks.  „„  Besides  the  four  events recorded by other indicators in the Mariposa  Grove,  only  two other  indicators  were  recorded  in the groves during the 20th  century.  Fire  scar  dates  were recorded  for  prescribed fires in Circle Meadow (1982) and in the Mariposa Grove  (1979).  The sharp  reduction in numbers of both fire scars and other indicators in the post-1850 period  rela tive  to earlier centuries also supports the contention that the other indicators reflected  primarily fire   occurrences  rather  than  non-fire  disturbances.  Non-fire  disturbances  that  might   have caused the other indicators, such as tree falls, lightning, etc. would have continued even  during the livestock grazing and fire suppression eras.  „„  Also  of interest was the 1875 fire recorded as fire scars on six trees in the  Atwell  grove. In  a  famous  passage  John  Muir  graphically described a fire  burning  into  this  grove  in  the "autumn" of 1875 (Muir 1901). The Visalia Weekly Delta of July 29, 1875 also reported: "Heavy  fires  are  raging  in the mountains east of here ... [giving  the  appearance  at  night]  of immense lanterns suspended from the heavens." (cited in Otter 1963). Another fire in 1864 that was recorded by our samples in the Mariposa Grove was referred to by Frederick  Law  Olmstead  in an address on August 8, 1865 before a  meeting  of  the  Yosemite  ð6 2  ŠValley  and Mariposa Big Tree Grove Commission (reference provided by Mr.  Stan  Hutchinson in October 2, 1989 letter to Nate Stephenson): "The  Commissioners  propose  also in laying out a road to the Mariposa Grove  that  it  shall  be carried  completely  around  it,  so as to offer a barrier of bare ground to  the  approach  of  fires, which  nearly  every  year sweep upon it from the adjoining country, and  which  during  the  last year  alone  [my  underlining]  have  caused injuries, exemption from  which,  it  will  be  thought before many years, would have been cheaply obtained at ten times the cost of the road." Another more direct report of the 1864 fire was published in the Mariposa Gazette vol. 10. no. 7, August 13, 1864: "Mariposa Grove. Editor Gayette - Sir - ... but on Wednesday last I visited the grove... Portions of the grove are now on fire. Several of the finest trees are suffering."  „„  Hartesveldt  (1964)  listed  an  1862 and 1889 fire in the Mariposa  Grove,  but  neither  of these  dates  were recorded by our fire scar specimens. However, it appears that  he  relied  on Presnall's  (1933a,  1933b) studies of fire scar dates that were determined by  increment  coring through  old  fire  scars on sequoia trees and counting the annual rings from bark  to  the  scars. Very  few  of  his  estimated  dates correspond with our exactly  dated  fire  scars  leading  us  to conclude  that  his dating was generally in error. The last fire scar we recorded in  the  pre-1900 period  was in 1864, and the only post 1900 scar was a 1979 scar caused by a prescribed  burn. Although  Presnall's  1862  date seems to be an error, his 1889 date may  be  correct.  Mr  Stan Hutchinson (pers. comm. to Nate Stephenson) stated that he had several references to a fire  in this  year  and that it burned in the southeast corner of the upper grove. However,  none  of  our sample trees recorded this fire.  ð6 2  Š „„  Barrett (1935) cited a reference to a fire in July and August of 1898: "A  fire set by sheep herders burned 20,000 acres of timberland on Marble Fork and  North  Fork of  Kaweah  River.  This  fire  badly burned many big trees  in  Sequoia  National  Park.  --  Ralph Hopkins, 1912". Barrett also cites another reference to an 1898 fire, perhaps the same one mentioned above: "August  26.  A forest fire is raging near Millwood in the mountains. It is reported that  the  fire  is spreading  toward  Big Trees in General Grant Park. The fire is beyond control.  San  Francisco Chronicle, August 27". Millwood was relatively close to the Big Stump Grove, but we did not identify any 1898 fire scars there.  However, Vankat (1977) suggested the 20,000 acre fire attributed to sheep herders  was the  same  "destructive fire [which] raged in Giant Forest" as reported the following  year  (1899) by  the  "Acting Superintendent". Although the Circle Meadow fire scar samples did  not  record an  1898  fire, several recently sampled and analyzed pine fire scar collections from  the  Colony Mill area below the Giant Forest did record a fire in this year.  „„  In  summary, fire frequencies based only on fire scars are certainly an  underestimate  of actual  fire  frequencies,  but they do provide a minimum estimate of fire  frequencies.  Fire  fre quencies  based  on  all fire indicators (fire scars and other indicators  combined)  are  probably closer to the true fire frequencies within the sampled areas, however, the precise errors of these estimates are unknown. Based on observations of a decline in both fire scars and other  indica tors  in  the  post-settlement period (after ca. 1850), and only a few  other  indicators  not  corre sponding  with known fire dates in this period, we estimate that less than 10% of the "fire"  dates recorded by other indicators may actually have been non-fire disturbances.  ð6 2  ŠRange and Variations in Fire Regimes  „„  Variations  in  the  distributions  of  fire intervals  were  evident  between  different  groves (Figs.  13  and  14). When all fires were considered the interval distributions  were  more  similar (Fig.  13)  than  the  interval distributions of fires recorded by two or  more  trees  (Fig.  14).  The latter  comparison excludes fires that may have been less extensive, and emphasizes the  inter vals between events when more area apparently burned.  „„  Our  ability  to  directly  compare reconstructed fire  regimes  from  different  groves  was limited  by  differences in the size of the area sampled and the numbers of sampled  trees.  Fire historians  have  long recognized that mean fire intervals are typically inversely related  to  study area  size (Romme 1980, Arno and Peterson 1983) because as more area is sampled the  likeli hood of including the burned areas of additional small or large fires increases. Even if the areas and numbers of trees sampled in each grove were the same, the spatial distribution of fuels and topographic  features in each grove were surely unique. Although sampled areas of the  groves differ  in  shape, and in other important ways, the total areas defined by the convex hulls  of  the five  sets  of sampled trees were not greatly different, i.e., they do not differ by orders  of  magni tude. The Mountain Home set was the most closely grouped with an estimated area of 13.2 ha. (Fig. 6), and the Mariposa set, including samples from both the "upper" and "lower" groves  was the most broadly distributed with an estimated area of 69.3 ha. (Fig. 3).  „„  The  occurrence  of  some longer fire-free intervals in the Circle  Meadow  and  Mountain Home  groves (Fig. 14) could be related to the relatively smaller areas included in  these  collec tions.  However,  both  of  these  areas were generally more  mesic  than  most  other  sites  we sampled,  and  so it seems just as likely that any differences in fire intervals may be  site  related rather than due to our sampling. We cannot visually detect any other clear differences  between the interval distributions among the groves.  „„  Despite  some  differences  in  sampled areas the mean fire  intervals  (MFIs)  in  the  five groves  for  the  period  A.D. 500 to 1900 were remarkably similar,  especially  for  the  estimates based  on  all fire indicators. Some differences between groves were observed when  only  fires  ð6 2 Š..Figs. 13-14, pgs. 34-35 .pn 36 recorded  by  fire  scars, and fires recorded by more than one tree  were  considered  (Table  4). About  3 to 4 year MFIs were recorded by all indicators, and 4 to 8 year MFIs were  recorded  by fire  scars.  About  5 to 10 year MFIs were recorded by all indicators when more  than  one  tree recorded  the  fire,  and about 5 to 16 year MFIs were recorded by fire  scars  alone  when  more than  one tree was scarred. The estimate based on fires scars alone was, as mentioned  earlier, an underestimate of true fire frequencies.  „„  It is possible that some fire scars represent more severe or intense fires, since they were fires that actually caused a lesion that we sampled. However, we doubt that most fire scars  are directly  related  to  fire  severity or intensity within a grove because  fire  scars  are  formed  very easily  on  trees  with previous scar cavities. These cavities tend to accumulate  litter  and  other fuels  between  fires, and these fuels may be ignited by very low  intensity  burns.  Furthermore, the exposed wood at the surface of older scars also can ignite from low intensity burns, and the protective  bark is often thinner at the edge of wound so new scars may be formed by  relatively low fire intensities and duration of heating. On the other hand we observed particular fire  years that were recorded by a majority or all of our sampled trees within groves, and sometimes these scars  encompassed an unusually large proportion of the circumference of the  partial  sections. We  also  noted  that some fire years consistently had greater  magnitude  and  duration  growth releases  following  them  than  other fire years (see later section on  the  Mountain  Home  1297 fire).  Investigation the sequoia fire record in relation to fire severity and intensity is an  objective of  future studies of the database. For now, the fire scar MFI estimates should be viewed as  the most conservative estimate of fire occurrence - i.e., these were the longest MFIs that could have occurred  in  our sampled areas in this time period with absolute confidence.  The  MFIs  based on  all  indicators  are more likely to be closer to the actual fire  occurrence  within  our  sampled areas.  The MFIs based on fires recorded by more than one tree should be viewed as an  esti mate of intervals between relatively widespread fires, or larger areas burned.  „„  It is likely that some fire dates recorded by more than one tree within each sampled area were  actually  separate  fires  that did not have continuous burned areas,  and  may  even  have  ð6 2 Š.pn 38 ..table 4, pg. 37 burned  at  different  times in the year. It is not possible to determine the  actual  extent  of  past fires  based on fire dates at discrete points. The best we can do is estimate relative  changes  in fire  frequency and extent estimated at points (single trees), and a composite of points (a  group of trees or site).  „„  Mean  fire intervals were useful as a general descriptor of fire regimes, but  maximum  or minimum  intervals between fires were probably more ecologically important. For example,  the shortest  and  longest  fire-free interval within an certain area during a certain  period  may  have been  crucial for determining whether or not individual plants were able to germinate and  estab lish.  Relatively long intervals without fire may be particularly important for some tree  seedlings to  develop  sufficient  height  and bark thickness to  withstand  surface  fires.  Conversely,  very short  intervals  between fires may favor other kinds of plants, such as grasses.  Short  intervals may also imply less intense and spatially patchy fires because of lower fuel accumulations,  and conversely,  long  intervals  without fires may be followed by relatively  intense  and  widespread fires.  In  either case, distinctly different ecological effects might result from  changes  in  maxi mum or minimum fire-free intervals.  „„  During the pre-settlement era maximum fire frequencies (all fire indicators) within groves were  usually  at  least double, and sometimes triple or greater  than  the  minimum  frequencies (Table  5,  and Appendix C). For example, from the period A.D. 500 to 1799 the  maximum  and minimum centennial fire frequencies in Circle Meadow were 42 fires in the 1200s and 12 fires  in the  1700s.  This  was equivalent to mean fire intervals (MFIs) of about 2.5 and  8.6  years.  The shortest  and longest MFIs in the other groves were: Mariposa - 2.6 years in the 1700s  and  7.8 years  in the 500s; Big Stump - 1.8 years in the 1600s and 6.1 years in the 600s; Atwell Mill -  2.2 years  in  the 1000s and 6.2 years in the 500s; Mountain Home - 2.9 years in the 1100s  and  7.6 years  in  the 1700s. Thus, similar ranges of centennial MFIs were observed in  the  groves  with longest and shortest MFIs differing by about 4 to 6 years.  „„  All  of  the  fire chronologies had a minimum fire intervals of 1 year  during  each  century from  A.D. 500 to 1799 (all fire indicators, Table 5). In the highest fire frequency  centuries  there  ð6 2 Š..Tab. 5, pgs. 39-43 .pn 44 were  many instances of consecutive year fires. For example, 21 different  consecutive-year  fire occurrences  were  recorded  during  the  1200s in  Circle  Meadow,  which  was  the  maximum recorded  centennial fire frequency in this area of the Giant Forest. In 19 cases  these  consecu tive year events were recorded on different trees, and in 14 cases they were recorded on  differ ent subgroups within Circle Meadow. Hence, these shortest interval fires were typically smaller, patchier  fires  that probably did not spread much beyond the size of a group  of  sampled  trees (<  1.0  ha.).   Although consecutive-year fires were usually recorded on different  trees,  we  did observe  consecutive  year fire scars recorded on different partial sections from  the  same  tree, and  rarely,  even on the same partial section. Thus, on rare occasions,  consecutive  year  fires occurred in the vicinity of a single tree.  „„  The  maximum  fire-free  intervals  varied more between groves. From  500  to  1800  the briefest  maximum  fire  interval  during  century-length periods was  4  years  in  the  Big  Stump Grove  during  the 1600s, and the longest was 30 years in the Mariposa Grove  during  the  600s (all  fire  indicators,  Table 5 and Appendix C). The 1800s and 1900s  typically  had  the  longest maximum  fire  intervals in the groves, reflecting the end of the episodic fire  regime  around  the time  of  Anglo-American settlement. The ranges of maximum fire intervals in  years,  computed each  century,  for  the  period  500  to  1800  were:  Mariposa,  6-30;  Big  Stump,  4-22;   Circle Meadow, 7-23; Atwell, 7-27, and Mountain Home, 6-18. Fire Size versus Frequency  „„  An  inverse  relation between disturbance size or severity and  disturbance  frequency  is implicit  in the concepts of disturbance regimes (Pickett and White 1985). However, we are  not aware  of  empirical  data that has clearly demonstrated this pattern  within  a  single  vegetation type  over  time.  Usually, the comparison is between types - for example,  high  severity  stand- replacing  fires  occurred very infrequently in boreal forest regions of Canada,  and  low  severity surface  fires  occurred  very  frequently  in  southwestern  pine  forests  in  pre-settlement  times (Heinselman 1981). (Note that the term "severity" is used in the qualitative sense here, in  refer ence  to  the  impact  of  fire on vegetation. The term "fire  intensity"  is  sometimes  used  inter ð6 2 Š changeably  with  "fire severity", but the term "intensity" has a connotation of heat  release  (e.g., kw/m2) because of a widely used quantitative term in fire management - "fireline intensity".  We generally use the term "intensity" because inferences regarding "high" or "low" intensities of past fires  (i.e.,  more or less heat) seem more justifiable than inferences regarding  ultimate  impacts on the vegetation.)  „„  In  analyzing  the  changes  in  fire  frequency within  the  sequoia  groves  we  noticed  a tendency  for  higher frequency periods to be dominated by apparently smaller fires,  and  lower frequency periods to have more widespread fires. Our estimates of sizes of fires was relatively crude - a simple percentage of trees recording fires. Nevertheless, given that our sample  areas were relatively small and homogeneous, we assert that the percentage of trees scarred by each fire  was a proxy for relative area burned within the sampled study area. (We plan to  investigate this  further  in future studies by computing the convex hull areas of individual fires  recorded  by the  sampled  trees.)  An  illustration  of an inverse fire  frequency/size  relationship  is  the  high frequency  fire period in the 1000-1300 period in Circle Meadow, which was  generally  recorded by  lower  percentages of trees than the low frequency period after 1300 (Fig. 10,  15).  A  com parison  of  fire  sizes, estimated by percentages of trees recording  fires,  versus  fire  frequency during  century-length  periods  shows  that a weak inverse pattern may  have  existed  in  some groves  but  not others (Fig. 16). The Atwell set appears to have the strongest  inverse  relation ship, while no clear association is visible in the Big Stump comparison. Temporal Trends in Fire Regimes Late 19th Century Fire Decline: The Complex of Ignition Source, Livestock Grazing and  Climate Change.  A  sharp decline in numbers of fires occurred in all groves in the late 1800s  (Fig.  17). The  mid-19th  century end of the surface fire regime in the Giant Forest is further  confirmed  by recent  collection  and  dating of fire-scarred pines on the long ridge  extending  from  the  south edge  of  the  grove  (near Huckleberry Meadow) to a point directly  above  and  south  of  Circle Meadow. This ridge is currently dominated by white fir and sugar pine. Fire scar dates in Circle  ð6 2 Š..Figs. 15-17, pgs. 46-48 .pn 49 Meadow  correspond  with fire dates from the ridge during the period from ca. 1700  to  present, although  a  few  more fires were recorded on the drier pine sites on the ridge. The  last  fire  re corded  in  Circle Meadow occurred in 1863, and the last fire recorded on the ridge  occurred  in 1873.  „„  This  decrease  in fire activity appears to be related to the settlement era  of  central  Cali fornia  and  the  Sierra  Nevada, beginning with intensive grazing by sheep  or  cattle  within  the groves,  and  subsequently  the organized suppression of fires by  government  authorities  who gained  jurisdiction  around  the  turn  of the century. Overgrazing  of  the  Sierra,  especially  by sheep, is well documented (Vankat 1977, Dilsaver and Tweed 1990). Drought conditions in  the early  1860s stimulated San Joaquin Valley ranchers to move their livestock into the  high  coun try.  The  abundant  grasses  they  found there encouraged  a  rapid  expansion  of  herds,  and subsequent denudation of formerly lush meadows and hillsides. By 1861, Hale Tharp - the  first white  inhabitant  of  the  Giant Forest - was grazing horses and cows in  this  area.  In  1864  as many  as  4,000 cattle were reported in the Big Meadows area (Vankat 1968),  which  is  roughly midway  between  the  Giant  Forest and Big Stump Grove (Fig.  1).  Sheep  grazing  intensified during  the early 1860s in the Mountain Home area with bands of 2,000 or more  sheep  grazing in  the  meadows near our sample trees (Otter 1963). The last fire scar at Mountain  Home  was recorded in 1864.  „„  We  observed a similar sudden elimination of surface fire regimes in  Southwestern  U.S. fire-scar  chronologies that closely followed the rise of livestock grazing, but  typically  preceded organized  fire  suppression  efforts  by  one to several  decades  (Swetnam  1990,  Savage  and Swetnam 1990). We suggest that grazing animals removed fine fuels that were important to fire spread  in  these  high frequency fire regimes. Moreover, trampling  of  ground  vegetation  and development of trails by livestock and herdsmen would tend to limit the spread of fires.  „„  An  alternative  explanation  advocated  by  Kilgore  and  Taylor  (1979)  was  that  Native Americans  were  an  important augmenting source of fire ignitions in  Sierra  Nevada  forests  in presettlement times. By the mid-1860s many of the tribes were forced out of the region or  were  ð6 2 Šdecimated  by  diseases  introduced  by  Europeans (Dilsaver  and  Tweed  1990).  Kilgore  and Taylor argued that fire frequencies remained high for one to several decades longer because  of range  burning by sheepmen. Although sheepherders were blamed for most fires in  the  Sierra Nevada  during  the  1870s  through  the turn of the century, it is  likely  that  their  influence  was overestimated  by  early  observers who did not appreciate the  high  frequency  of  lightning-set fires  (Vankat  1977, 1985). For example, note that the fire attributed to sheepmen in  1898  (see quote  on page 22) was in July and August, yet sheepmen usually set fires in the fall  when  they were  leaving  the  mountains  (Barrett 1935, Otter 1963). If  the  majority  of  written  statements regarding  the  prevalence  of fires started by sheepmen were entirely true we  would  expect  an increase,  or  at  least  a continued high frequency of fires  through  the  1890s  when  significant government  control  began  to be exercised in the National Parks (Dilsaver  and  Tweed  1990). However,  Kilgore and Taylor's fire history generally shows a decline of fires by the  1870s.  Our recent  collections  of fire scarred samples from pine stands within and below  the  Giant  Forest show  a  reduced  fire  frequency after about 1863, and effectively, an  end  to  episodic  fires  by 1898.  „„  From  fire suppression records Kilgore and Taylor counted 15 lightning-started fires  be tween 1921 and 1975 in one of their watersheds (Redwood Creek). This would equate to a  MFI of about 3.6 years, but they stated this was not a sufficient frequency to account for their recon structed  fire regime. This argument rested largely upon the assumption that  lightning-set  fires would  not have spread to large areas because fuels would have been too low  and  discontinu ous in the pre-1875 forest. Citing ethnographic evidence of the widespread use of fire by Native Americans in California (Lewis 1973, Reynolds 1959), Kilgore and Taylor concluded that Indians must  have greatly supplemented lightning-set fires in pre-settlement sequoia-mixed conifer  fire regimes.  Curiously,  although they discounted the total number of observed  lightning-set  fires within  their  watershed because they did not believe most of them could have spread  into  their fire-scar clusters, they later suggested that Indian set fires:  ð6 2  Š"-  either  in the coniferous forest itself or as escapes from lower-elevation  woodland,  grass,  or chaparral  fires  -  augmented  the lightning-caused ignitions to  give  the  frequencies  we  have found." Similarly, Kilgore and Taylor quoted Bonnicksen (1975): "Most  descriptions  of Indian burning were recorded from low elevation areas  and  these  were usually  in grassland or brush .... There is a possibility that many, if not most Indian-caused  fires were the result of escapes from low-elevation areas."  „„  It seems reasonable that if Indian-caused fires could have spread to the sequoia  groves from  relatively  distant  low elevation areas, then lightning-caused fires could have  done  so  as well.  Summaries  of lightning fire records since the 1920s in Sequoia National  Park  present  a striking  picture of the high density of lightning fire occurrence, even in the lower elevation  foot hills  zone  (Parsons  1981, Vankat 1985). From the Sequoia and Kings Canyon  fire  data  base (1930-1986) we counted a total of 28 lightning fires, 90 human-caused fires, 8 prescribed  burns and 7 fires of unknown cause in the nine sections centered on the Giant Forest. These sections were  T15S,  R30E,  Secs.  31, 32; T16S, R30E, Secs. 5,6,7,8; T16S, R29E,  Secs.  1,  12;  T15S, R29E,  Sec.  36. This approximate 2,300 ha area encompasses virtually all of  the  "bench"  that the  Giant  Forest  rests upon, plus part of the steep slopes to the south and west  of  the  bench above  the  Middle  and  Marble Forks of the Kaweah River. Given  that  significant  numbers  of lightning  fires  start  as  early  as  June and July, and  dry  conditions  usually  continue  at  least through  September,  there  would  have been plenty of time for  fires  to  spread  from  relatively distant  locations  into  the sequoia groves. It is reasonable to infer  that  some  lightning-ignited fires  in this area would have spread into the Circle Meadow area of the Giant Forest.  Hence,  if only  half  of these 20th century lightning fires (which were suppressed)  had  eventually  burned into Circle Meadow, the MFI for this period (1930-1986) would have been about 4.1 years. Even  ð6 2 Šif  we  consider only the four sections directly centered on the Giant Forest (Secs. 6, 5,  31,  and 32)  with  Circle  Meadow  almost directly in the middle of this area, the  total  count  of  lightning ignited  fires  was 18 during this 57-year period. If only half of these fires made it into  the  Circle Meadow area the MFI would have been 6.3 years.  „„  Since A.D. 1500 the estimated MFIs in Circle Meadow were: 1500s = 6.2 years, 1600s = 4.9 years, 1700s = 8.6 years, 1800s = 18.5 years, 1900s = 56.5 years (see Table 5 and Appen dix  C).  Therefore, we conclude that there was a sufficient amount of lightning  fire  ignitions  to account  for  all, or nearly all fires that occurred in the Circle Meadow area in  the  pre-settlement era back at least to A.D. 1500.  „„  From  A.D.  800 to 1500 we reconstructed centennial MFIs with a range from  2.5  to  3.9 years  in  Circle  Meadow. If we follow the assumption that ignition rates  by  lightning  were  not substantially  different  from the twentieth century rates, and only half or fewer of  all  ignitions  in the  nine  sections  around Circle Meadow would be likely to spread  into  Circle  Meadow,  then observed  fire  intervals  may  be  too short to be accounted  for  solely  by  lightning.  However, assumptions  about  long-term  changes in lightning ignition rates and fire  spread  patterns  are very  tenuous  because  these  topics  have  not been studied  in  much  detail  by  anyone.  We propose  that  climate  change  is an equally valid, if not  more  compelling  explanation  for  ob served  changes  in fire frequency than explanations based on changes  in  human  populations and  burning  practices.  A  higher frequency of droughts, or  other  "fire-inducing"  climate  may have  occurred  during the higher fire frequency periods. Different climate  conditions  probably have  different  rates of lightning occurrence (e.g., see Price and Rind 1991).  Climate  changes could  also influence fire spread patterns through changes in fuel production,  types,  condition, and  distribution  on the landscape. Hence, both the number and proportion of  fires  ignited  at distant locations successfully burning into the groves could have changed through time.  „„  While  the  ethnographic  evidence  appears  strong  that  Indians  used  fire  extensively, especially  in  the lower foothill and valley areas of the Sierra Nevada, it is possible  that  pre-set tlement  fire frequencies in sequoia-mixed conifer forests would have been the same,  or  nearly  ð6 2 Šthe  same,  if  Indians had not contributed at all. The important point is that, at  least  during  the past  four or five hundred years, ignition sources in this zone were probably not  limiting.  Simi larly,  fires  set by sheepherders may have been less of an ecological impact than the  effects  of grazing  and  trampling  of  the  fine fuels. We argue that  fuel  types,  amounts,  condition  (i.e., moisture content) and spatial configuration were the controlling features of this system.  „„  An  important qualification to this argument, however, concerns the actual timing of  fires that were set by people versus lightning. Many references state that Indians and  sheepherders used fire late in the season, i.e., late summer or fall (Barrett 1935, Otter 1963, Lewis 1973, Wick strom  1987).  This generally corresponded with the lightning fire season. For  example,  light ning fires starting in July, August and September accounted for 37%, 31%, and 18% (86%  total) of  the total annual occurrence in Sequoia and Kings Canyon National Parks  (1930-1986).  The peak  times of burning would probably have been in late August or September when  fuel  mois tures are usually lowest (Parsons 1981). Thus, Indian or sheepherder-set fires would  ordinarily mimic  the seasonal timing of lightning fires. However, if people consistently set fires  earlier  or later  than  July  through September, then the contribution of human-set fires  could  have  been ecologically  important. Furthermore, these fires could have been particularly  important  during some  drought  years  if  rates of lightning occurrence were lower  due  to  sparse  thunderstorm activity.  In either case, with relatively high frequency fire regimes prevailing during  the  pre-set tlement  era,  fuel  amounts, conditions and distribution would  necessarily  have  been  primary factors determining fire spread and severity.  „„  Obviously,  the  topics of grazing impacts, fire sources, and climate  variation  are  com plex,  and  deserving  of further investigation and discussion.  Investigation  of  lightning  occur rence  patterns in the vicinity of the other four groves (Mariposa, Big Stump, Atwell,  and  Moun tain Home) is needed. We are planning a more thorough review of this topic for presentation  in later papers.  ð6 2  ŠSynchronous  and  Non-synchronous  Fire Occurrence: Centennial-Scales.  From  A.D.  500  to present  there  was  a general similarity in decadal to centennial-scale temporal  changes  in  fire frequency  among  the  groves, with some important grove-specific  differences  overlaying  this similarity.  Fire  frequencies were consistently low from about 500 to 900 and then a  rise  in  fire frequency  occurred  from about 900 through the 1200s. Most of the chronologies  displayed  a sharp decrease in fire occurrence in the middle of the 1300s. From the late-1300s to the end  of the surface fire regime in the mid-1800s there was a general decrease in fires, except for rises in fire  occurrence  in  the 1600s in Big Stump, and in the 1700s in Mariposa and  Atwell  (Figs.  17, 18).  „„  The  lower fire frequencies before ca. A.D. 1000 might appear to be related to  a  general decrease in the number of samples (see Fig. 11). However, there were relatively small  declines in  sample  numbers from Mountain Home and Circle Meadow during  several  centuries  before 1000 and relatively low fire frequencies were also recorded in these two groves. Even within the other  groves,  where sample depths were declining through this period, we  believe  the  overall record  was  fairly  well documented with a minimum of 6 trees recording scars by  600,  and  10 trees by 800.  „„  Another possible bias in the fire record could be a decrease in numbers of observed  fire indicators in earlier periods because of poor preservation of older scars and wood, or an inabili ty  to sample older fire indicators deeply embedded in huge sequoia stems. Although the  latter problem  may  have been an important bias in some of the earliest periods  sampled  within  the groves (e.g., B.C. periods) we doubt that it was generally a problem in most individual trees  we sampled  because  we  often  were able to obtain partial sections from  the  center,  or  near  the center  of  large sequoias. Deep fire scar cavities enabled us to cut samples from  these  areas. Also, most of the v-cut radial sections from stumps included the pith, or came close to it  (within one  or two centuries). Dating and observations of the radials provided records of non-fire  scar indicators  (e.g., growth releases and traumatic resin ducts). Furthermore, we found that  pres ervation  of  completely  healed-over  fire  scars in the  partial  sections  was  generally  excellent  ð6 2 Š..Fig. 18, pg. 55 .pn 56 regardless  of  the age of the material we sampled. Many partial sections had clusters  of  10  or more  healed  over  scars within a relatively small area, and some samples had as  many  as  35 scars in compact, healed-over fire scar clusters. Altogether, embedded, healed-over fire  scars were  the  most  common  fire scar type within samples and they generally  showed  little  or  no evidence  of  loss of the fire record due to decay or "burning off" of the earlier fire scars  by  later fires.  „„  Although  we  have not yet devised a quantitative test of the possible  effect  of  temporal changes  in  the  quality  of  the fire record within individual sample  trees  (as  discussed  in  the previous  paragraph) we have analyzed the effect of changing numbers of trees sampled  within groves  on  the  fire  frequency estimates. We simulated reduced sample  sizes  in  each  of  the groves  and  recalculated  the fire frequencies. The period A.D. 1000 to 1599 was  chosen  as  a base  period  because  during  this  time (1) maximum  sample  sizes  (numbers  of  trees)  were achieved  in all groves, (2) sample sizes decreased or increased by only one to three trees,  and (3)  this  period  included both high and low fire occurrence, and relatively large  and  small  fires (inferred  from  percentages of trees scarred per fire). Thus, although this base period  was  still somewhat  arbitrary,  it was a reasonable period for comparing fire frequencies during  times  of smaller sample sizes and differing fire regimes.  „„  The  simulated mean fire frequencies at reduced sample sizes had a curvi-linear  decline from  the  maximum sample size to the sample size of a single tree (Fig. 19). The curves  for  all fires  appeared  to  be  ascending  to an asymptote, but  none  were  exactly  level  between  the maximum  sample  size  (right hand side of x-axis) and reduced sample  sizes.  Theoretically,  if the  complete  fire  record  was obtained by a reduced sample size,  and  the  effective  sampled area  was not reduced with the decreased sample size, then the curve should level out  (i.e.,  an asymptote) before reaching the maximum sample size. The amount of "effective" area sampled per tree was dependent on fire size. For example, if most fires occurring within the sample area were  large relative to the size of the sample area, one would expect relatively few  sample  trees would be necessary to record a complete or nearly complete fire history for the sample area.  ð6 2  Š.. Fig. 19, pg. 57 .pn 58 Alternatively,  if  many  fires  within  the sample area were very small  relative  to  the  size  of  the sample  area, then a relatively dense network of sampled trees would be necessary to  obtain  a complete  fire history. The minimum density of this network, plus one sample tree,  would  have to  be  such  that  removal  of one tree from the sample set would  not  affect  the  fire  frequency estimates because at least one other tree would have been close enough to record all fires  that the  removed  tree  had recorded. In our case, the amount of effective  area  sampled  probably changed  with  sample size because many of the fires occurring in the  groves  were  apparently small and may have burned only around a single tree or small group of trees. Also, fire  records from  individual  trees  were  probably not complete, i.e., not all  fires  that  burned  around  their bases  were  recorded  in our samples. Thus, in practice, it is probably  not  possible  to  obtain such  a  sufficiently  dense  sample network within the sequoia groves  to  obtain  an  absolutely complete  fire  record.  Although  we may have been approaching  minimum  sample  sizes  for obtaining  a nearly "complete" record, the lack of an asymptote (all fires, Fig. 19)  suggests  that the fire record was incomplete.  „„  It  appears that most (if not all) undetected fires were small fires that usually burned  only around  one  or  a  few  trees. This is supported by simulations  of  fire  frequencies  at  reduced sample  sizes  when  only  fires recorded by more than one tree were  considered  (Fig.  19).  In these  cases,  the curves for all groves were very nearly asymptotic  before  reaching  maximum sample  sizes.  Fire  frequency estimates differed by less than one fire  per  century  at  reduced sample  sizes of 28 to 42 percent (5 to 8 fewer trees included) of the maximum sample  sizes  in the  five groves (Fig. 19). This suggests that the record of fires large enough to be  recorded  by more than one tree was complete, or nearly so.  „„  For  both cases - i.e., all fires and fires recorded by two or more trees - we assessed  the relative changes in fire frequencies through time by comparing the observed fire frequencies  to the  estimated  fire  frequencies for the base period of 1000-1599 at reduced  sample  sizes  (Fig 20).  Observed  centennial  fire frequencies, plotted at their  corresponding  sample  sizes  were sometimes  higher  and  sometimes  lower  than  the curves  estimated  for  the  base  period  at  ð6 2 Šreduced  sample  sizes  (Fig. 20). The number of occurrences of observed  centennial  fire  fre quencies  above  and below the estimated fire frequency curves in all five groves  were  counted and  plotted  as  fire frequency "trends" in Figure 21. These graphs  showed  that  the  medieval period (ca. A.D. 1000 to 1300) was the most consistently high fire occurrence period during  the past  1,500  to  2,000  years. Also, fire frequencies during the period from  ca.  700  through  the 800s,  and  the  1900s were consistently lower than the base period fire  frequency  adjusted  for sample  size.  Almost all other centuries also typically had lower fire frequencies than  the  base period, but there was considerable grove-to-grove variability (Fig. 21).  „„  The  observed centennial fire frequencies based on all fires were generally closer  to  the mean fire frequency curves than the observed fire frequencies based on fires recorded by  more than  one  tree (compare upper and lower plots in Fig. 20). All of the observed  fire  frequencies based on fires recorded by more than one tree were outside of the 95% confidence limits of  the mean  curve  (above or below) (see lower plot Fig. 21). This appears to be due to  the  fact  that the 1000 to 1599 base period was overall a high fire frequency period, with only a few  individual centuries  in  some  groves having much higher frequencies during  this  period  (1000s,  1100s, and 1200s) (Fig. 21).  „„  In  summary,  the period from about 500 to 1300 was characterized by  a  relatively  syn chronous  change  in centennial fire frequencies among the groves, with an  approximate  dou bling  of the fire frequency from the earlier (ca. 500-900) to the later period (ca. 900 to  1300).  A general  decline in fire frequencies occurred in the groves after 1300, but this decline  was  inter rupted  by non-synchronous increases in fire frequency in the 1600s or 1700s. All of the  groves show  a  sharp  drop in fire frequency after about 1860. The fire  history  record  is  conservative within  the  groves, i.e., not all fires that occurred within the sampled areas were  recorded,  par ticularly  the  smaller  fires. Sample sizes did affect fire frequency  estimates,  but  the  observed centennial  fire  frequency  changes  mentioned  above  were  still  present  when  the  effects  of sample size changes were considered.  ð6 2  Š..Figs. 20,21, pgs. 60,61 .pn 62 Synchronous  and  Non-synchronous  Fire  Occurrence:  Annual-Scales.  Another  measure  of similarity and dissimilarity of fire occurrence among the five groves was provided by tallying  the number  of groves that recorded each fire date. Years with fires occurring in the majority  (ò  3) of  the  groves  might be regionally important fire years, whereas years with fires  occurring  in  a minority  of  the  groves (1 to 2) might be locally important fire years. Years  when  no  fires  oc curred in any of the groves might also be regionally important non-fire years. Regionally  impor tant  fire  and  non-fire  years were more likely to be controlled  by  climatic  patterns  than  other years because climate is the most logical exogenous factor that could cause such synchrony of events among widely distributed sites.  „„  Fires  were  highly  synchronous among the five groves during some  periods  (Table  6). The  highest  synchrony periods were from ca. 750 to 1300, and the late 1400s  to  early  1500s. The  Medieval  period from ca. 950 to 1300 had the highest synchrony of fire  events.  A  maxi mum  of  9  five-grove and 12 four-grove fire events were observed during the  1100s  (Table  6). Similar  temporal patterns were also observed when only fire dates recorded by fire  scars  were analyzed.  „„  The highest fire synchrony during the Medieval period corresponded with maximum  fire frequencies during this period (Fig. 18). However, the high synchrony does not appear to be an artifact  of the high fire frequency during this period. From simple probability statistics we  esti mate  that  the  joint  probability  of fire occurring in all five groves  in  the  same  year  during  all periods was less than 0.01. In other words, less than 1 fire event per century would be  expect ed  to  co-occur  in all five groves by chance (Table 7). There were only  four  century  or  longer periods  between  500 and 1900 when there was not at least one five-grove fire event  (Table  6). These periods were 500 to 651, 699 to 809, 1263 to 1492, and 1654 to 1795.  „„  A  much  higher  than  expected number of four-grove fire  events  also  occurred  during almost all periods, but especially during the 700s to 1300s, and during the 1500s (Table 7). The observed number of one and two grove events was much lower than expected from the proba bility statistics, and the zero grove events were consistently much more common than expected  ð6 2 Š..Tabs. 6, 7, pgs. 63, 64 .pn 65 by chance (Table 7). The Chi-square statistic was computed using the expected and  observed numbers of events per century. It was necessary to combine the 3, 4 and 5 grove categories so that  expected  cell  frequencies  were  greater  than 5.0  -  a  necessary  condition  for  this  test. Expected  number  of events in the 3 to 5 grove combined category was still only 3 for  the  19th century,  hence  the  Chi-square  value  computed for this  period  is  suspect.  The  Chi-square values  generally  indicate  that  there  was significant synchrony (p  <  0.05)  in  fire  occurrence during  nearly  all  centuries.  Curiously,  the 700s had the highest  Chi-square  value,  but  no  5 grove fire events. This appears to be due to high synchrony of both the non-fire years (0 groves recording a fire) and much higher than expected occurrence of 3 and 4 grove fire events  during this century.  „„  In  summary,  the significant synchrony of fire chronologies among  the  sequoia  groves was  strongly indicative of a fire-climate response in sequoia-mixed conifer forests.  Local-scale (grove)  differences  in the fire regimes were also important, but a part of the  annual  to  century scale variability must be linked to regional-scale climate variations. Seasonal Timing of Fires  „„  Of  a total 3,243 fire dates we were able to identify the relative position within  the  annual ring  for  2,705,  i.e.,  83% of the total (Table 8). These fire dates were the sum  of  all  fire  dates recorded by trees in the 5 groves. The actual number of observations of fire scars examined for all  of these dates was at least several times higher, because the individual and multiple  speci mens  from  the  same tree often had the same fire date replicated several or  more  times.  The "unidentified"  portion  of the examined scars was greatest among the  Mariposa  grove  speci mens (28%) and lowest among the Circle Meadow specimens (7%). There was also variability in  the  size  of  the unidentified portion through time within each  grove  (Table  8).  There  were many potential sources of this variability, including growth rates (e.g., scars within smaller  rings were  more  difficult  to classify), amount of decay, weathering or  charring,  and  observer  bias.  ð6 2  Š..Tab. 8, pgs. 66-70 .pn 71 The  latter  source, observer bias, was expected since these identifications were visual  and  re quired  some subjective judgment. However, we instituted a set of standard identification  crite ria fairly earlier in our laboratory analysis, and members of our dating team interacted closely on difficult  classifications. In informal tests we found that agreement on relative scar  position  was high  when  two different trained workers independently examined  specimens.  Altogether,  we judge that growth rates and state of preservation of specimens were the most important  factors influencing our ability to consistently classify the relative positions of scars.  „„  A  substantial majority of all examined fire scars occurred within the latewood  portion  of the  annual  ring  (Fig. 22). The highest percentage of latewood scars was in  Big  Stump  (91%) and  the  lowest was in Circle Meadow (46%). Among all five groves the mean  percentages  by position  were:  earlywood 0.6%, early-earlywood 0.8%,  mid-earlywood  10.2%,  late-earlywood 12.2%,  latewood  66.4%, and dormant 9.8%. The "earlywood" category was  specified  when  it was  not  possible to determine precisely which part of the earlywood the scar was  within.  The other  three  earlywood categories refer to the first, second and third "thirds"  of  the  earlywood. The  "dormant"  category  was used when the scar occurred along the boundary  of  two  annual rings,  i.e.,  with  latewood cells on the "pith side" of the scar and earlywood  cells  on  the  "bark side" of the scar (see Appendix A).  „„  Temporal  variability  in the relative scar positions was also apparent (Tables  8  and  Fig. 23).  A  couple  of  subtle  patterns were visible in the area  graphs  that  illustrate  the  temporal changes  in  scar  position. After ca. A.D. 1300 there was a slight increase in  the  proportion  of dormant  season  scars,  and an increase in latewood scars in the  two  southernmost  groves  - Atwell  and Mountain Home. No latewood scars were identified in the 3 fire scar dates  in  Circle Meadow during the 1800s (Fig. 23). Many other fluctuations in these graphs appeared to  show variations  that  were  unique  to  individual groves.  Big  Stump  was  particularly  unusual,  with latewood scars dominating throughout the record much more than in the other groves.  „„  The  significance  of  these  patterns  and trends is not entirely clear  to  us  at  this  time. Interpretation  of  scar  position is complicated by changeable seasonal  "windows"  of  cambial  ð6 2 Š..Figs. 22, 23, pgs. 72, 73 .pn 74 growth and fire occurrence. Elevation and other site-related differences, and year-to-year  varia tions in weather certainly lead to differences in the timing of the onset of growth in the  summer, rates  of growth, and cessation of growth at the end of summer or in the fall. The seasonality  of fire is also affected by these factors.  „„  Despite  these  complications, we believe that some useful  interpretations  on  seasonal timing  of  past fires from fire scar observations are possible. Since 1988  scientists  at  Sequoia National  Park (Dr. David Parsons and others) have been monitoring cambial growth  of  mature sequoias  in  the  Giant Forest. Increment cores taken at about weekly  intervals  from  six  trees were  used  to determine approximate onset of earlywood and latewood growth, rate  of  growth through the season, and cessation of growth in the late summer or early fall. Preliminary compi lations  of four years of observation (1988 to 1991) show that growth initiated by late June  in  all years. Typical detection dates of growth initiation of earlywood among the sampled trees were: June 20, 1988; June 23, 1989; June 18, 1990; June 26, 1991. Typical detection dates of growth initiation  for  latewood were: Sept. 16, 1988; Sept. 8, 1989; Sept. 12, 1990; Oct. 23,  1991  (data from  D.  Parsons).  Cessation of growth was more difficult to identify with  certainty.  A  lack  of change  in total ring width after latewood formation had begun would generally be taken  as  the approximate end of the growing. This appears to have occurred fairly soon after latewood  initi ation in most years - approximately within one to three weeks. The actual initiation dates of  the earlywood  and  latewood type cells could have been a week or so earlier than  these  estimates because of the week-long observation intervals, and because observation of the newest formed tissue  is  generally not possible on increment cores that are simply surfaced with a razor  or  by sanding with abrasives. Only lignified cells are clearly visible on such specimens. Thin sections mounted on slides are usually necessary to see newly formed unlignified cells.  „„  From  these  observations  we could infer that earlywood-type fire  scars  were  probably produced  by  fires  burning between the first week of June and the last week of  August  or  first week of September. Latewood-type fires were probably formed by fires occurring from the  first week  of  September to third or fourth week of October. Dormant season  scars  were  probably  ð6 2 Šformed by fires after mid-October, but before temperature and precipitation changes associated with  fall  and  winter greatly reduced the chances of fire spread. This  would  be  approximately early  to mid November, depending on the year. There is a smaller chance that  some  dormant season scars were formed by spring fires, i.e., before cambial growth initiated in June. Howev er, less than 5% of lightning ignited fires started before June in this century (1930-1986), and fire spread during this time is usually inhibited by high fuel moisture at these elevations.  „„  Although we believe these inferences are generally correct, there are reasons to be cau tious  in  immediately  accepting them as broadly representative of seasonal timing  of  fires  re corded  by  the  fire scars. The first is that all of the years of observation  of  cambial  phenology are within one of the droughtiest periods in California in many decades. Therefore, it is possible that  the  initiation,  rate, and cessation of cambial growth during  this  episode  are  anomalous. Variability  in the "window" of the cambial growing season is illustrated by the  surprising  obser vation  of  initiation  of  latewood formation in 1991 in late October, more  than  one  month  later than observed in previous years.  „„  We  conclude that latewood fire scars, representing the largest proportion of scars  in  all five  groves, were probably formed by fires occurring after late August and before  late  October. However,  more observations of cambial phenology in years of differing climatic  conditions  are necessary  to  obtain  a  broader estimate of the variability in  the  phenological  window  of  se quoias.  We  believe  that a useful future approach would be to employ a  new  cambial  growth model  (Fritts  et  al. 1991) that simulates cellular growth of trees given  daily  weather  variables. The soil moisture budget and photosynthesis are included in this model. When a version of this simulation model is parametrized for sequoias it should be possible to explore different  climatic conditions  that  would  change  the  cambial  phenology  window.  NPS  scientists  (Drs.   Mark Finney  and  David Parsons) are currently conducting experiments on marking the  cambium  of sequoias with pins at different times of the year, and artificially inducing "fire" scars. Information from  these  studies  will be of great value in improving our interpretation  of  seasonal  timing  of past fires from fire scar observations.  ð6 2  Š A.D. 1297 Fire and Growth Release  „„  An  unusual  fire  event  in the year A.D. 1297 in the Mountain Home  grove  is  a  striking illustration  of  the wide range of possible fire regimes in giant sequoia forests. Nine of  15  trees recorded  a fire scar at this date, and all 15 trees had another fire indicator at this date. All  trees showed a tremendous increase in growth following 1297 (Fig. 24). The magnitude and duration of  this release, and the consistency of this anomaly among sampled trees, was not equaled  by any other growth release in the five groves we have intensively studied. However, increment core  collections  from  trees in the Redwood Mountain Grove show a very  strong  release  after 1297,  and  a  fire  scar was dated at 1297 on one tree. Additional  sampling  will  be  needed  to determine if a similar event occurred there.  „„  The  maximum ring growth during the release period at Mountain Home was  often  over 200%  of  the 100-year average ring growth prior to the fire. The growth release was  so  visually obvious, even on unsanded surfaces, that we were able to identify it on many stump tops in  the Mountain Home Grove in locations separated by several kilometers. Additionally, we noted that among the 16 sequoia stumps that Douglass sampled and dated from Mountain Home, eight of nine  trees  that  established  before 1200 exhibit the post-1297 growth  release,  and  six  of  the seven  remaining  trees apparently germinated within 20 years following the  1297  event.  Pith dates  of  the  younger  trees  varied from 1300 to  1327.  Considering  that  most  of  Douglass' specimens  came from positions at least one meter above ground level, it is likely that  germina tion  dates  closely followed the 1297 event. Two of our sampled fire-scarred trees (MHF  1  and MHF40) also appear to have established during this period (Table 3).  „„  We  hypothesize  that the post-1297 growth release was the result of a  substantial  thin ning by fire, i.e., fire-caused mortality of many trees. We also note that pre-1297 growth rates in some  samples  were markedly low with little ring-width variation (see MHF 4B in Fig.  24).  This suggests  that  these  trees  were in a competitive situation of a closed  canopy  forest  that  was limiting  growth.  The  many large stumps remaining from the harvesting era of  the  1880s  and  ð6 2 Š..Fig. 24, pg. 77 .pn 78 1890s,  and the still living mature sequoias at this site indicate that many sequoias  survived  the 1297  fire. Thus, one possible scenario would be an intense fire in a fairly dense stand of  trees. The  fire  may  have killed a large proportion of understory trees and possibly  also  some  larger sequoias. Mortality of trees and opening of the forest canopy would be consistent with both the observed growth release of survivors and the establishment of another cohort of sequoias.  „„  We  are  currently investigating the 1297 fire by mapping locations of  stumps  exhibiting the  release, and by studies of sequoia growth response in prescribed burns and  historic  local ized high severity burns (Swetnam 1991b, Mutch and Swetnam 1992). Studies on sequoia  tree age  structure  in  the  Giant Forest are also being conducted by  Dr.  Nate  Stephenson.  These studies  are within fire scar collection sites. Preliminary observations show that sequoia  cohort establishment  has generally followed surface fires during the past several  centuries  (Stephen son et al. 1991). .pa  2 Š ½½! SUMMARY, CONCLUSIONS, AND MANAGEMENT IMPLICATIONS  „„  Some  major  findings  of  this study are summarized below in  boldface  type.  We  offer interpretations  and  recommendations  deriving from these results that  have  relevance  to  the future management of the sequoia groves. These comments represent the views of the investi gators and not necessarily those of the National Park Service. Giant sequoia tree rings contain unique and valuable histories of environmental changes.  „„  The temporal length and high resolution of the fire history records contained in sequoias are  unmatched  in  any  other tree species or forest ecosystem yet  studied.  We  were  able  to reconstruct  fire  regimes  back  to at least A.D. 500 in five sequoia  groves.  Several  of  the  fire chronologies  also  contained  useful information well back into the B.C. period.  In  addition  to exact  dates  (years)  of  past  fires we were able to  document  the  approximate  season  of  fire occurrence.  „„  Other  valuable  paleorecords are being obtained from sequoia tree rings,  such  as  cli mate  history  (Hughes  and Brown 1992) and sequoia age structure (Stephenson  et  al.  1991). Studies  of  intra-ring  structure,  density,  and isotopic  composition  of  sequoia  tree  rings  are currently  underway (M. K. Hughes pers. comm.). These records and the fire  chronologies  are likely  to play a significant role in national and international efforts to understand past and  future global  changes  because they provide a rare source of information on  environmental  changes on seasonal to millennial time scales. The unique insights and historical perspectives  provided by  these data offer managers and scientists a solid basis for evaluating the past role of fire  and other  natural  processes,  such  as climate, in ecosystem dynamics.  These  data  also  offer  a tremendous  educational  opportunity. Interpretive materials relating  information  derived  from tree-ring  studies  to the public can greatly enrich the visitor's experience and  understanding  of sequoias.  „„  There  is considerable potential to obtain additional information from sequoia tree  rings, and  to  extend and expand fire history and climate studies to other groves that  have  not  been  ð6 2 Š.. .pn 79 studied yet. Of particular value are the remnant sequoia snags, stumps and logs present in cut- over and virgin sequoia groves. Future studies of subfossil sequoias buried in meadow or  river sediments  may also provide even longer records of environmental change. These dead  mate rials  can  be  considered  "natural archives" with unique and  irreplaceable  histories.  We  urge managers to consider the scientific values of remnant sequoia material before actions are taken that  would impact this resource. Prescribed burning, road building, and harvesting of  windfalls or  old  sequoia stems on the forest floors are some ways that this resource could  be  gradually lost. Evaluation of such resources in areas to be impacted, and in some cases "salvage"  sam pling  of  sequoia  tree rings, may be appropriate mitigation measures (Van  Pelt  and  Swetnam 1990). Fire frequency, size and severity within the groves were highly variable through time.   „„  Maximum  fire frequencies within sampled areas of sequoia groves were as high as 3  to 4  fires per decade. Lower fire occurrence periods had 1 to 2 fires per decade. Occasional  fire- free intervals lasted 20 to 30 years. Generally, centennial maximum and minimum fire  frequen cies  in the five groves since A.D. 500 differed by a factor of 2 to 3. Many consecutive-year  fires were  recorded  during the highest fire frequency periods within sampled areas,  but  these  fires were probably "patchy" and burned only small areas - possibly less than 1.0 ha., or the size of a small  group  of  sequoias. Periods with longer intervals between fires usually  had  more  wide spread fires. Some of these fires may have burned through the majority of the grove.  „„  Because  of  the relatively high fire frequencies during almost all time periods,  fuel  load ings  were usually low and most fires must have been relatively low intensity surface  fires.  Fuel studies  have  shown that litterfall accumulates relatively quickly in  Sierra  mixed-conifer  forests (Agee  et al. 1977, Parsons 1978). Using a simulation model (FYRCYCL) that incorporated  fuel dynamics  and  weather  components in Sierra Nevada mixed conifer  van  Wagtendonk  (1972) found  mean  fire intervals of about 4 years. However, because our fire  history  reconstructions identify some periods with even more frequent fires (i.e., 1 to 3 year mean fire intervals) it seems possible that changes in fuel production rates or types of fuels may have been important in past  ð6 2 Šfire  regimes.  For example, it is possible that the understory of  pre-settlement  sequoia  groves had  a  much larger component of grasses than are present today,  especially  during  extended periods  of higher fire frequencies. Presence of grasses would help explain the occurrence of  1 to  3  year  interval fires in the groves, and the virtual elimination of spreading fires  at  about  the same  time  that intensive livestock grazing began in the 19th century.  Unfortunately,  no  clear descriptions  of pre-1860s understory vegetation or photographs of pristine,  ungrazed  sequoia groves  are  available. Alternatively, fuel types may have changed little, but production  rates  of needles  and  woody  fuels  from trees may have been high enough  to  support  the  higher  fire frequencies.  „„  Occasional  higher  intensity  surface  fires, or localized crownfires,  may  also  have  oc curred  after  extended  periods (> 10 years) without fire when branches, tree  boles,  and  other fuels accumulated. Some of these fires probably caused mortality of small groups of  understo ry and overstory trees. This pattern may explain observations of clusters or groups of  relatively even-aged  sequoias within the groves that established in fire-created openings (Stephenson  et al.  1991).  Sequoia  seed dispersal from partially serotinous cones also seems to  be  linked  to this  pattern since high intensity fires tend to result in greater amounts of seed released and  low intensity fires result in lower amounts of seed released (Rundel 1992).  „„  Large,  high  intensity  fires,  such as the 1297 fire in Mountain  Home,  may  have  swept through  sequoia  groves on very rare occasions. We speculate that this fire killed a  large  pro portion  of  understory  trees and possibly some monarch sequoias as well.  A  large  cohort  of sequoias seems to have established following this fire. Although it is too early to conclude  with certainty  that  such  widespread high intensity fires also occurred  in  other  groves,  preliminary observations  from  the  Redwood  Mountain Grove suggests that a similar  event  in  1297  may have  occurred  there.  Regardless  of the rarity of this type of fire, our  discovery  of  it  certainly expands the window of possible fire regimes that sequoia groves can sustain.  „„  We  hypothesize that the end of the episodic fire regime in the late 19th century was  pri marily due to the introduction of intensive livestock grazing, and subsequently, fire  suppression  ð6 2 Šby  government  agencies.  It  is very likely that some fires in the  sequoia  groves  were  set  by Native Americans. These fires could have been set within the groves, or they may have  burned into  the groves from lower elevations. However, we argue that ignition sources  were  generally not a limiting factor in sequoia fire regimes, before or after departure of the local tribes of  Native Americans,  because  rates  of  lightning  ignitions were sufficient  to  account  for  observed  fire frequencies.  We suggest that fuel types, amounts, condition, and spatial distribution  were  the determining factors in regulating past fire regimes.  „„  Regardless of the relative importance of Indian versus lightning ignitions, or the  ultimate causes of the 19th century fire decline, it is clear that the past century of fire-exclusion was  truly anomalous.  No  other fire-free period in the previous 2,000+ year fire history was even  half  as long. Furthermore, sequoia seedling and age structure studies clearly show that fire is and  has been  a  key  component of sequoia regeneration for centuries  (Hartesveldt  and  Harvey  1967, Harvey  et  al.  1980,  Stephenson et al. 1991). Hence, reintroduction of fire  to  the  groves  is  a highly  logical  and eminently justifiable objective. However, prescribed  burning  programs  will eventually  need  to more specifically address some basic questions regarding  the  means  and methods  of  restoring  the balance of fuels, fires and tree population  processes  (i.e.,  mortality and  regeneration).  Of  particular importance will be decisions  regarding  appropriate  burning intervals, fire sizes, and fire intensities.  „„  One  implication  of  the  high temporal variability of the historic  fire  record  is  that  past sequoia  fire  regimes  cannot  be characterized by a single fire frequency  or  mean  fire  interval estimate.  In  determining  appropriate  burning intervals managers should  not  establish  a  set interval; the range and variability of past fire occurrence must be explicitly recognized.  Manag ers  wishing to replicate or simulate past fire frequencies within groves have a very broad  range of  surface  fire regimes to choose from. This does not make the choice easy. What  is  the  fire regime  that  should be chosen? If we use the past as a guide, should we  reintroduce  fire  fre quencies  that  were  sustained during the 600 years before the settlement era,  when  10  to  20 year  fire-free  intervals  were relatively common, or should it be the fire regime  of  the  Medieval  ð6 2 Šperiod  when  fires  often occurred every other year and no fire-free  intervals  were  longer  than about 5 years? Because tree regeneration and mortality patterns are so intimately linked to  the fire frequency/intensity complex these decisions could have profound long-term consequences on  sequoia  grove  structure. No doubt, fire exclusion in the past  century  has  already  greatly altered the structure of sequoia groves and the surrounding forests (Parsons and  DeBennedetti 1979,  Kilgore  and Taylor 1979). In most cases it will probably not be practical  to  simply  allow "mother  nature"  to decide these issues by letting lightning ignited fires to burn  within  and  into the groves. The landscape is now fragmented by roads, trails, and dwellings, and fuel structure and  arrangement  has changed too much to assume that such fires would burn as  they  did  in the past.  „„  Control  of  fire  sizes  and fire intensities in sequoia groves are  also,  ultimately,  the  re sponsibility  of  fire  managers. Perhaps one day some of these fire regime  factors  may  be  al lowed  to develop entirely "naturally" from fires that are simply ignited within groves,  or  allowed to  burn  from  lightning  fires started within or outside of  groves.  However,  fuel  loadings  and arrangements  are too high in many places to follow an exclusively laissez faire burning  strate gy.  Until  the  fuel  loadings within groves are restored to levels that  are  more  similar  to  what existed  during  pre-settlement  times  -  a dynamic  controlled  by  plant  productivity  rates  and consumption  by fires - managers will have to continue existing practices of controlling  sizes  of burn  units  and  fire intensities by dispersing or isolating fuels in some areas,  and  by  following refined burning prescriptions.  „„  Ultimately,  scientists  and managers must determine what sort of fire regime is  likely  to perpetuate  sequoia  groves for an indefinite time into the future. The past fire history  is  only  a partial  guide  to the solution of this problem. The fire history helps to define  the  "envelope"  or boundaries  of  fire  regimes  that the sequoias have  developed  within.  More  information  and tools  are needed to determine fire spread and intensity patterns given different fire  frequencies and  fuel  loads.  The  short  and  long-term effects of  these  different  fire  regimes  on  species composition and age structure of sequoia mixed-conifer forests should be studied. Monitoring,  ð6 2 Šexperiments and simulation modeling efforts currently supported by the Sequoia, Kings Canyon and Yosemite National Parks and the NPS Global Change program will hopefully provide  some of this information. Most  pre-settlement fires were recorded as latewood-type fire scars indicating  most  area burned in late summer to early fall.  „„  We  were  able to determine the intra-ring position of 83% of all of the fire  scars  we  ob served in the five groves. The intra-ring position of the remainder could not be clearly  identified because of narrowness of rings or because of decay or other problems. The successfully  iden tified  positions  represent  thousands of observations of fire scars. A large  majority  of  the  fire scars  were within the latewood (about 66%). About 22% occurred from the middle one-third  of the  earlywood to latter one-third of the earlywood. About 10% of the scars were  dormant  type scars. Although the majority of intra-ring positions in all groves were of the latewood type, there were  some  differences  in proportions among the groves and through time  within  the  groves. Big  Stump  consistently had the largest proportion of latewood-type scars  (>90%),  and  Circle Meadow had the lowest (about 43%).  „„  From  recent  observations of cambial phenology in sequoia groves  (D.  Parsons,  pers. comm.)  we  tentatively  interpret these scar positions to represent fires that  burned  during  the following approximate times of the year: latewood-type fires burned from about the first week of September  to  the third or fourth week of October; earlywood-type fires burned from  about  the first  week  of  June  to the last week of August or first  week  of  September;  dormant-type  fires burned after mid-October, but before mid-November.  „„  Because  the  cambial phenology measurements of sequoias were conducted  during  a drought  period in California it is possible that our interpretations are biased by an  unusual  pat tern  of  cambial  growth. Additional measurements and experiments are  being  conducted  by NPS  scientists  (Drs.  David  Parsons and Mark Finney) that should  help  to  further  clarify  this issue.  With  this new knowledge, it is possible that our interpretation of the seasonal  timing  of past fires will shift by one to several weeks.  ð6 2  Š „„  Despite  the uncertainty associated with these interpretations, it is clear that the  majority of area burned late in the growing season of sequoias. Even though we will need some refine ment  of our understanding of cambial phenology, a confident interpretation from  existing  data is  that  the peak burning season of most fires in sequoia groves (latewood-type fires)  was  after the  first  of  August and before the end of October. Because the timing of  burning  is  a  critical factor  in encouraging or discouraging some plant species, park or wilderness  managers  wish ing  to  simulate  pre-settlement  vegetation patterns should  carry  out  some  prescribed  burns during  this period. Obviously, consideration of the hazards of escaped fires and undesired  fire intensities during this normally hot and dry period must play an important role in burning  plans. This  will be especially true until fuels within the groves have been restored to lower levels  more closely approximating pre-settlement conditions. However, the long-term goal of  reintroducing some  early  season  fires  (June  to August) and a larger proportion  of  late  season  fires  (after August) should be an ultimate goal of some sequoia grove fire management plans. Fire  occurrence  among  the  five  groves  was  partially  synchronized  on  centennial  and annual  time  scales, indicating that regional climate variations were important  to  sequoia fire regimes.  „„  Similar  patterns  of  high  and  low  fire occurrence  among  the  groves  since  A.D.  500 suggest  that  regional  climate patterns have been partly responsible for  variations  in  past  fire regimes.  The  most consistent patterns appear to be relatively low fire activity from  about  A.D. 500  to  900  (especially during the 700s and 800s), followed by an increase in  fire  activity  from about 900 to 1300 (especially during the 1000s through 1200s). A general decline in fire activity occurs  in  the  groves  after 1300, especially in the mid-1300s.  However,  other  increases  that were not synchronous among the groves occurred in the 1600s and 1700s.  „„  Temporal  changes in sample sizes affect fire frequency estimates, and  therefore  com plicate  the  interpretations of fire frequency changes. However, in our analysis  that  accounted for  changes in sample sizes we still found similar patterns of highest fire frequencies during  the medieval period (about 1000 to 1300) and lower fire frequencies before and after this time.  ð6 2  Š „„  We  hypothesize that the consistent trends toward higher fire frequency in  the  medieval period  were  related  to a generally drier climatic period in the  Sierra  Nevada.  Tree-ring  width chronologies  and  precipitation reconstructions from foxtail pine, western  juniper  and  sequoia generally  support this hypotheses (Graumlich 1991, Swetnam et al. in prep.). This  period  was known  as  the Medieval Warm Epoch in Europe. It was generally a time of glacier  recession  in the  Alps,  lower  amounts of sea ice in the North Atlantic, and  favorable  crop  production  and tree-growth in some northern European countries (Lamb 1977).  „„  The strongest evidence for a regional-scale climate influence on the sequoia fire regimes is  the  remarkable synchrony of fire dates and non-fire dates among the five groves.  Nearly  all time  periods  had  a  much higher coincidence of common fire  years  and  non-fire  years  than would  be  expected to occur by chance. One of the highest synchrony  periods  also  occurred during  the  Medieval Warm Epoch. This suggests that, not only was this a time of  frequent  fire inducing climate (e.g., droughts), but these climate patterns were regional in scale resulting in a highly synchronized fire occurrence in the Sierra Nevada.  „„  These  observations  suggest  that  sequoia fire regimes were  influenced  by  very  large scale - perhaps global - climate patterns. This means that sequoia grove structure and compo sition  has  been at least partly regulated by exogenous factors. The degree to  which  endoge nous  or  local  factors,  such  as intra and inter-species competition,  and  site  factors  such  as topography  and soils, have dominated over the exogenous factors, or vice versa,  is  unknown. This  is a key question and research challenge for ecologists wishing to understand  the  devel opment  and dynamics of ecosystems (Ricklefs 1987). The implications for  ecologically-orient ed  management  of  these ecosystems are profound. For example, if  sequoia  groves  were  a product of predictable (i.e, repeatable) patterns of fire, tree mortality, tree regeneration, etc., in a dynamic equilibrium controlled by internal factors, then a continuous quasi-cyclical schedule  of burning  could  perpetuate  the groves in a state that is similar to past  structures  and  composi tions.  However,  if  sequoia  groves were a product of  unique,  aperiodic  external  forces  (i.e.,  ð6 2  Šclimatic  events  or trends) then the structure and compositions of sequoia groves at  any  given point in time are unique, and basically un-repeatable.  „„  We  believe  that the latter case is more likely than the former. What does this  mean  for management  of sequoia groves? For one thing, it means that we must accept that  multi-direc tional  change - rather than cyclical change - is the historical (natural?) pattern  for  ecosystems. Hence,  perpetuation  of ever-unchanging sequoia groves would be something new  and  artifi cial.  Although  the National Park's policies now are to reintroduce  natural  processes  (roughly defined as pre-settlement historical processes), the great emotional attachment of the American people  to sequoias may ultimately force managers to attempt to perpetuate sequoia  groves  in some  kind of "desired state", by whatever means are possible. If predicted changes  in  climate due  to  increasing  greenhouse gases do occur, then this task may be  forced  upon  managers sooner,  rather  than later. Difficult questions raised by this possibility include: What is  the  de sired  state of sequoia groves? And perhaps more important: Is it really possible  and  practical to perpetuate sequoias in a desired state in the face of environmental changes? We are skepti cal  about  our ability, or the desirability, of fully controlling the future of sequoias in  a  changing world. However, we believe that reintroduction of fire to the sequoia groves is both an  ecologi cal imperative and an opportunity to mitigate negative impacts of human-caused changes in the environment.