The study was conducted along two similar drainages, the North Fork of the Humboldt River (T45N R41E, Sec. 19) and Cabin Creek (T44N R40E, Sec. 5), within the Santa Rosa Mountains in northern Nevada, in the western USA. Elevation of the two drainages was approximately 1800 m. Woody vegetation was predominantly Artemisia tridentata and Chrysothamnus nauseosus and C. viscidiflorus in the uplands, grading into low-stature riparian vegetation (mainly Salix spp.) near the rivers. Juncus balticus, Carex microptera, C. nebraskensis and C. praegracilis, and Poa pratensis were most common in the riparian zones; predominant upland species included Poa secunda and Agropyron trachycaulum (Table 1). Soils were loamy, with gravel inclusions on stream terraces. Average annual precipitation is 30.8 cm (12 in.); the frost-free season averages 80 days.

Description of Chemicals

We used one Class A (i.e. applied to Class A fuels, such as wood) fire suppressant foam, Silv-Ex*, and one fire retardant, Phos-Chek G75-F, in our field tests. Silv-Ex concentrate is a proprietary mixture of sodium and ammonium salts of fatty alcohol ether sulfates, higher alcohols, and water, as well as butyl carbitol and ethyl alcohol (Ansul, Incorporated 1994). It functions as a surfactant (i.e. detergent), allowing water to penetrate and expand over the surface of fuels to both cool and smother the fire (Pyne 1984). Silv-Ex, like other Class A foams, is applied operationally either from ground tankers or helicopters. Silv-Ex is supplied by the manufacturer as a liquid concentrate, which is mixed with water to the desired concentration before application.

Phos-Chek G75-F is a proprietary formulation composed of monoammonium phosphate and ammonium sulfate, fugitive coloring agent, and small amounts of gum-thickener, bactericide, and corrosion inhibitor (National Wildfire Coordinating Group, Fire Equipment Working Team 1991). Phos-Chek is typically applied from helicopter bucket or ground tanker in advance of a fire; other retardants with higher viscosity are applied from fixed-wing aircraft. The ammonium salts retard fire by chemically combining with cellulose as fuels are heated, as well as through evaporative cooling of the fuels (Lyons 1970, Pyne 1984). Phos-Chek is supplied by the manufacturer as a powder, which is mixed with water to the desired concentration before application.

Plot-based Treatments

Treatments were applied to plots in a stratified random design, divided equally between riparian and upland habitats and between North Fork and Cabin Creek drainages. Treatments included: (1) Phos-Chek; (2) Phos-Chek/burned; (3) 0.5% Silv-Ex; (4) 0.5% Silv-Ex/burned; (5) 1% Silv-Ex; (6) 1% Silv-Ex/burned; (7) water, equivalent in volume to chemical application; (8) water/burned. Because each chemical was mixed with water, water was used on the control plots so we could distinguish the effects of added moisture from the effects of the chemicals in this moisture-limited environment. We applied the treatments to 1-m² plots, which is an appropriate size for vegetation analysis in the habitats under investigation (Bonham 1989); because wildfire was of concern in this arid region, the small plots also afforded us better control of experimental burns. Cattle exclosures measuring 1 m² × 1 m high were placed around each plot. Exclosures were made of 6-cm woven wire fencing and were anchored with steel rods. Each of the eight treatments was applied to each of four randomly located plots in the riparian zone and four randomly located plots in the upland zone on 28 June–1 July 1994. The procedure was repeated on different plots on 19–20 July to determine the effect of time of application within the growing season on vegetation response.

Chemicals were mixed with water as appropriate for operational use on sagebrush communities. Phos-Chek was applied at coverage level 3 (115 L/ha). Silv-Ex was mixed at two concentrations, 0.5% and 1.0%, and applied at the rate of 1410 L/ha. We used motorized 25.37-L backpack pumps to apply the chemicals. We did not quantify expansion of the foam. When treatment included burning, plots were ignited with a propane torch; all vegetation within the plot was burned, although not all was reduced to ash. Chemicals or water were applied to extinguish the fire, depending on treatment.

Prior to treatment, we marked five individual shoots (terminal segment of a branch) on each Chrysothamnus viscidiflorus within the upland vegetation sampling plots. We measured current year’s growth (leaders) of these shoots, and counted the number of stems (ramets) and the number of post-burn resprouts of Chrysothamnus (total number of new ramets sprouting from each marked Chrysothamnus plant after treatment). We counted number of stems and number of species on each plot and calculated species diversity (H’) and evenness (J’) for each plot (Ludwig and Reynolds 1988). Total live stems on burned plots and total species on all plots were counted within 2 weeks before treatments were applied (pre-treatment), at 4, 6, 8 and 13 weeks after June applications, and at 2, 4, 6, and 11 weeks after July applications. Total live stems were counted on unburned plots before treatment and at the last two sampling periods after treatment. The last sampling period corresponded to the end of the growing season at our study sites.

Statistical Analysis for Plot-based Treatments

We used analysis of variance (ANOVA) techniques to assess the effects of the treatments on the change in number of stems, plant species richness, diversity (H’), and evenness (J’; where species richness > 1), from pre-treatment to post-treatment sampling periods. We conducted separate ANOVAs for each habitat type (riparian, upland) because species varied between them; separate analyses were also performed for the June and July applications because of phenological differences among plant species. The ANOVA model was of the form:

y_ijkl = + _i + _j + _ij + _k(ij) + _l + _il + _jl + _ijl + _ijkl,

We used ANOVA techniques to assess the effects of the four burn treatments (water/burn, 0.5% Silv-Ex/burn, 1.0% Silv-Ex/ burn, and Phos-Chek/burn) on total Chrysothamnus viscidiflorus plants per plot. Separate ANOVAs were done for June and July applications. For number of plants per plot, the ANOVA was a simple one-way in a randomized block design with location (North Fork, Cabin Creek) considered as random blocks. The ANOVA was a simple one-way with subsampling (Steel and Torrie 1980) for stems/plant and stem length; individual plants within plots were the subsamples, with a mean stem length estimated for each plant within each plot. Only plots that had C. viscidiflorus prior to treatments were used in the ANOVAs for number of stems per plant and stem length. Sample size for these variables ranged from 2 to 4. Because we had few plots in which we were able to assess treatment effects on stems per plant and stem length, we pooled plots across the two locations for these response variables.

We used canonical variate analysis (CVA) (Jongman et al. 1987, Gittins 1985), also called linear discriminant ordination (Green 1979, Pielou 1984), to compare the vegetative community among treatments and examine shifts in the community across sampling periods. We used CVA as an ordination procedure because it permits the differences among treatment groups and sampling periods to be displayed with maximum separation (Pielou 1984:235). We were able to use CVA as an ordination technique because the number of plots sampled exceeded substantially the number of species encountered (Jongman et al. 1987:149). We excluded from the CVA rare species (i.e., species occurring in fewer than 5% of the plots; Gauch 1982:214). Separate CVAs were done for each habitat zone (riparian and upland) and for each application time (June and July). We used Pearson correlation coefficients to correlate the canonical variates with each of the species abundances to interpret the canonical variates and describe the type of community they reflect. All species abundance values were ln(y+1) transformed prior to the CVA. We used the Canonical Discriminant procedure (PROC CANDISC) of SAS (SAS 1989) to conduct the CVA.

Big Sagebrush (Artemisia tridentata) Response

Non-burn treatments (i.e. 0.5% Silv-Ex, 1.0% Silv-Ex, Phos-Chek, and water) were applied to 40 randomly selected Artemisia plants (not located on vegetation plots), 10 per treatment. The volume of chemical applied to each plant was scaled according to the approximate volume of the plant, using 1.41 L/m³. We applied the chemical using the same motorized backpack pumps we used for the vegetation plots.

We randomly selected and tagged four branches on each plant. Current annual growth (terminal leader and four subsequent leaders), and number of galls per branch were recorded prior to treatment and at the end of the growing season, 5–6 October. When flowers were present, we measured inflorescence length at the end of the season.

We used separate one-way ANOVAs to test for pre-treatment differences among the four groups in plant height, plant volume, and amount of solution to be applied. We used repeated measures ANOVA (Diggle et al. 1994) to assess differences among the four treatments and any change over time from pre-treatment to post-treatment with respect to mean leader length and mean number of galls adjusted for leader length. Mean leader length and mean number of galls used in the repeated measures ANOVAs were first averaged across the five leaders from each of four branches for each of the 40 plants. A one-way ANOVA was also used to test for differences among the four non-burned treatments with respect to inflorescence length at the end of the season. A mean inflorescence length was computed for each plant by averaging across individual flowers and across the four branches.

We used the general linear models procedure (PROC GLM) of SAS (SAS 1989) to conduct all ANOVAs except the mixed model described above. For all ANOVAs, we used Fisher’s protected LSD test to isolate differences among least squares means for significant main effects or significant interaction effects, if any (Milliken and Johnson 1984). Significance was set at P = 0.05.