Spring and summer (April to July) precipitation varied from 101 mm to 325 mm (Figure 1), characteristic of long-term variation in growing season precipitation during 1939 to 2016 (mean = 196 mm, standard deviation = 63 mm, range = 53 mm to 453 mm). Annual variation in April to July precipitation was strongly correlated with mean soil VWC during May to July (r2 = 0.92, P < 0.001), which in turn was a primary driver of annual variation in herbaceous plant production (r2 = 0.53, P = 0.04).
Standing Dead Biomass, Bare Soil Exposure, and Soil Moisture
In all years except 2013, annual fire treatments reduced standing dead biomass by an average of 93 % (Figure 2a and 3a). Significant fire effects on standing dead biomass and bare soil exposure were clear throughout this experiment (year × treatment: F32, 120 = 13.54, P < 0.001; year: F8,120 = 26.82, P < 0.001; treatment: F4,15 = 26.85, P < 0.001; Figures 3 and 4). Triennial burn treatments significantly reduced standing dead biomass in the first post-burn year (average reduction: 90%), with the exception of the triennial autumn burn in 2013. Two years following the triennial treatment, standing dead biomass was still lower in the 2011 autumn triennial treatment (50%) and the 2016 spring triennial treatment (40%) relative to the unburned plots. Standing dead biomass did not differ between the third year post-burn plots and unburned controls. Due to low fuel loads and patchy burns, burning did not affect standing dead biomass in 2013. From 2008 to 2016, annual burns increased bare soil exposure by an average of 177%. Following 2008, the triennial spring fire treatment increased bare soil exposure relative to unburned controls every year except 2013. The mean increase in bare soil exposure in the first, second, and third years post treatment for the triennial spring burns was 116%, 145%, and 70 % of the unburned treatment, respectively. From 2009 to 2016, triennial autumn burns increased bare soil exposure in the first, second, and third years post treatment by averages of 139%, 152%, and 72%, respectively.
For soil volumetric water content (0 cm to10 cm layer) we observed a treatment × season interaction (F14, 203 = 374, P = 0.009). Burning had no effect on soil moisture in either spring (P > 0.74) or autumn (P > 0.24). In the summer, there was a significant treatment effect (F4,15 = 5.92, P = 0.005), with both annual autumn and triennial autumn burns having significantly lower soil moisture than control plots (Figures 2c and 3c). Averaged across years, mean summer soil VWC was 13.3 % in unburned plots, 11.8% in annual autumn burn plots, and 12.1 % in triennial autumn burn plots.
Productivity of Plant Functional Groups
The most productive plant during our study was the annual grass, sixweeks fescue. Biomass averaged only 1.8 g m−2 (range: 0 g m−2 to 2.8 g m−2) in the unburned treatment during 2006 to 2012, but then increased dramatically in abundance to an average of 26.4 g m−2 (range: 6.4 g m−2 to 45.9 g m−2) during the four-year period from 2013 to 2016 (Figures 4a and 5a). Both spring and autumn annual burning treatments reduced sixweeks fescue biomass during 2014 to 2016 by an average of 87% (Figures 4a and 5a). During the same time period, a reduction of similar magnitude (86%) occurred in the autumn triennial treatment, which was burned prior to the 2013 and 2016 growing seasons, relative to unburned controls (Figure 4a). In contrast, we found no significant effect of the spring triennial burn treatment, which was burned prior to the 2012 and 2015 growing seasons, on sixweeks fescue production relative to unburned controls (Figure 5a). This latter result was not surprising during the years of 2013 and 2014. However, even though the spring triennial plots were burned in spring of 2015, sixweeks fescue production was not suppressed in this treatment either in 2015 or 2016 (Figure 5a).
For C4 perennial grass production, burn effects varied among years and changed in direction of effect from the first year (2006) as compared to the last two years of the experiment (2015 and 2016; year x treatment interaction, F40,150 = 2.21, P = 0.0003; Figures 4b and 5b). We tested for contrasts between the unburned controls and each of the four burn treatments in those years with significant (P < 0.05) yearly effects. Spring burns significantly reduced C4 grass production in the first year of the experiment, likely because burns occurred after C4 growth was initiated (Scheintaub et al. 2009 and Figure 5b). However, plots that burned a second time during the dormant season in 2007 recovered to the same level of C4 grass production as unburned plots (Schientaub et al. 2009). From 2008 through 2014, none of the burn treatments affected C4 grass production (Figures 4b and 5b), despite the fact that precipitation varied widely among these years, from a drought in 2012 to substantially above-average precipitation in 2009. During the 2015 and 2016 growing seasons, however, annual and triennial autumn burning, which suppressed sixweeks fescue (Figures 3a and 4a), also increased C4 grass productivity (Figures 4b and 5b).
For C3 graminoid production, we detected a marginally significant year × treatment interaction (F40,149 = 1.36, P = 0.096), which led us to investigate individual year slices for potential treatment effects in one or more years. This analysis did not reveal any significant differences between burn treatments and unburned controls in any given year (P > 0.13 for all years). For forb production, we found no significant year × treatment interaction (F40,150 = 1.24, P = 0.18) and no main effect of burn treatments (F4,15 = 0.26, P = 0.89). Thus, fire effects on C3 graminoids and forbs were consistently absent across a wide range of precipitation conditions, varying from drought to substantially above-average precipitation.
Finally, for total forage production (the sum of forbs plus C3 and C4 perennial graminoid production, but excluding subshrubs and annual grasses), effects of burn treatments varied among years (year x treatment: F40, 150 = 2.34, P < 0.0001). Parallel with the results for C4 grass production, burns in the first spring of the study (2006) negatively affected total forage production (Scheintaub et al. 2009; Figure 5c). During the subsequent eight-year period from 2007 to 2013, none of the burn treatments significantly affected total forage production across a wide range of precipitation conditions. In 2014, total forage production on the triennial spring burn treatment (third year post treatment) was 28% lower than production in the unburned and the annual spring burns (Figure 5c), while autumn burn treatments showed no significant differences (Figure 4c). Following the burns in spring of 2015, total forage production in the triennial spring burn treatment returned to a similar level as unburned controls. Additionally in 2015, there was no significant difference in total forage production on any treatment compared to unburned controls despite increases in C4 grass production on both spring and autumn annual burns and triennial autumn burns. In 2016, all treatments that suppressed sixweeks fescue also enhanced forage production relative to unburned controls.
Time since Burn Analysis
When examining results from 2013 to 2016 in terms of time since fire application (instead of season and frequency of fire), we found that time since burning significantly influenced sixweeks fescue production (main effect, F2,69 = 25.11, P < 0.001), with the magnitude of effect varying by year (year × time since burn interaction: F5,69 = 3.95, P = 0.003; Figure 6). Plots burned in the current year consistently had less sixweeks fescue than plots that remained unburned for >2 yr (Figure 6). Plots burned in the previous year produced more sixweeks fescue than plots burned in the current-year plots in 2013 and 2016, but not in 2014 (Figure 6).
Relationship between Sixweeks Fescue and Total Forage Production
In the last three years of our study, six-weeks fescue production increased dramatically in the unburned and triennial spring burn treatments (Figure 5a). Furthermore, we detected the first significant effect of burn treatments on total forage production compared to the unburned controls since the first year of the study (Figure 4c and 5c). In each of these three years (2014 to 2016), variation in six-weeks fescue production across all plots in the study was inversely related to total forage production (Figure 7). In particular, total forage production was consistently low in all plots in which sixweeks fescue production exceeded 30 g m−2 (Figure 7).