1988 growing season. Two sites are grassland habi- 

 tat types and two are forest habitat types. 



Site 1 is a rough fescue/bluebunch wheatgrass 

 (Festuca scabrellalAgropyron spicatum) habitat type 

 (h.t.) (Mueggler and Stewart 1980) at 3,200 ft eleva- 

 tion with a northwest aspect. It was used as cattle 

 pasture through the 1950's but since then has been 

 grazed only by wildlife. Site 2 is Idaho fescue/ 

 bluebimch wheatgrass h.t.-westem needlegrass 

 phase (Festuca idahoensislAgropyron spicatum- 

 Stipa occidentalis). At 4,200 ft the aspect is south 

 to southwest. Site 2 was grazed by cattle through 

 the 1970's but now is grazed only by wildlife, with 

 significant elk use in winter and spring. 



Sites 3 and 4 are adjacently located in a clearcut 

 that was logged and dozer-piled in the mid-1960's. 

 The habitat type (Pfister and others 1977) is ponde- 

 rosa pine/snowberry (Pinus ponderosal 

 Symphoricarpos albus), although much of the larger 

 surrounding area is Douglas-fir/snowberry (Pseudo- 

 tsuga menziesiil Symphoricarpos albus). At 4,050 ft 

 elevation, the aspect of site 3 is south to southwest, 

 while site 4 is almost flat. The logging returned 

 sites 3 and 4 to an earlier serai stage, with 

 Richardson's needlegrass (Stipa richardsonii Link) 

 and rough fescue (Festuca scabrella Torr.) being 

 the most abundant grasses. The location was being 

 grazed by cattle and wildlife. We erected a fence in 

 the summer of 1988 to exclude the cattle but not 

 deer or elk fi-om sites 3 and 4. Deer and elk use 

 on these sites is relatively light. 



The soils at sites 1 and 2 are MoUisols. The site 1 

 soil is a well-drained Typic ArgiboroU with a grav- 

 elly loam texture. The site 2 soil is a Typic Haplox- 

 eroU gravelly loam (Bigarm series). The soil at sites 

 3 and 4 is an Alfisol — a Typic Eutroboralf silt loam 

 (Greenough series). 



Canopy cover and frequency of occurrence by spe- 

 cies were determined by cover microplot method 

 (Daubenmire 1959). Each replicated treatment plot 

 (y2o-acre) had five random transects, each transect 

 had five permanently marked microplot locations 

 (10- by 20-inch Daubenmire frames). For each 

 microplot we determined the cover value of every 

 species. Pretreatment plant community data were 

 collected for the 1988 warm season (July-September) 

 and the 1989 cool season (May-June). 



Herbicides were applied with a carbon dioxide 

 pressure-regulated backpack research sprayer in 



1989 starting after the cool season pretreatment 

 readings for each site. "Early" herbicide treatments 

 were made when spotted knapweed was in the ro- 

 sette to early bolt stage. "Late" treatments were 

 made when knapweed was in the early to mid-flower 

 stage (table 1). Herbicides were picloram (Tordon 

 22K) at 0.25-lb active ingredient per acre (a.e. per 

 acre), clopyralid (Stinger or Transline) at 0.25 lb a.e. 



Table 1 — Herbicide application dates 





cdiiy 



1 9tO 





ircallTl^nio 



ireaiinenis 



1 



5/25/89 



7/20/89 



2 



6/07/89 



7/21/89 



3 



6/21/89 



8/08/89 



4 



6/21/89 



8/07/89 



per acre, and a mixture of 2,4-D at 1 lb a.e. plus clo- 

 pyrahd at 0.19 lb a.e. per acre (Curtail). Each of four 

 sites contained six treatments and an untreated check 

 plot with three randomized blocks or rephcations. 



Postspray data collection began 1 year after the 

 1989 herbicide applications. We reread each micro- 

 plot twice a year — during the cool season (May- 

 June) and again in the warm season (July-August). 

 We have completed the 1990 and 1991 readings and 

 report them here. We will determine commxmity 

 composition through at least 1992. 



Plant community data summaries, including aver- 

 age species richness, total number of species, and 

 Shannon- Weaver indices, have been calculated by 

 treatment (Keane and others 1990) through the 

 1991 warm season. The statistical model is a form 

 of randomized complete blocks with repeated meas- 

 ures in time of the same transect/microplots (Winer 

 1971). 



We determined soil herbicide residues for three 

 depth increments — to 5, 5 to 25, and 25 to 50 cm — 

 at 30 days and 1 year after spraying. A fourth 

 increment — 50 to 100 cm — was sampled for picloram 

 at 1 year. We also measvu-ed picloram residues at 2 

 years in the to 5 and 5 to 25 cm depth increments. 

 We collected 12 subsamples for each combination of 

 sampUng date-site-treatment-depth increment. The 

 like subsamples were mixed to form a composite 

 field sample. The composite samples (two spht rep- 

 hcates each) were fi"ozen. The residue analyses were 

 conducted at the Agricultural Experiment Station 

 Analytical Laboratory at Montana State University, 

 Bozeman. 



We used quality assiirance and quality control 

 (QA/QC) procedures for both the ecological field 

 measurements and the herbicide residue analysis 

 work. We read randomly selected microplots a sec- 

 ond time for canopy cover at the end of each sam- 

 pling period for each site. The second reading was 

 blind; that is, the evaluator did not refer to the data 

 from the first reading. We combined data from the 

 blind microplots to form synthetic transects, then 

 we calculated diversity values. Diversity values for 

 synthetic transect were also calculated fi*om the 

 original readings of the same set of microplots. We 

 quantified the precision of commvmity measurements 



2 



