4 Discussion
TODO:
- Is the regularization imposed on our random effects for sites appropriate if the distribution of site effect is bi-modal because of north vs south facing aspects?
- Aspect is an important predictor that was not explicitly accounted for in our model, and may have led to inflated uncertainty
- Cost of treatments
- 1-hr redwood fuels differed from other studies in that we implemented a cutoff
- Discuss the decision to allow surface fuels to extend to greater than 2 meters
- It may have been better to analyze sprout composition directly using the binned data for the smaller sprouts, rather than converting to basal area based on an assumed midpoint diameter.
- Discuss the decision to combine live and dead vegetation fuels
- Discuss the decision to combine herbaceous and woody vegetation fuels
- Discuss the decision to combine duff and litter
- Discuss the decision to use basal area for redwood and tanoak and counts for Douglas-fir
- Discuss the similarity between vegetation fuel difference post-PCT and vegetation fuel loading pre-PCT
- Discuss the significance of 1-hr fuel loading pre-PCT vs post-PCT
- Discuss the significance of 10-hr fuel loading pre-PCT vs post-PCT
- Discuss the significance of 100-hr fuel loading pre-PCT vs post-PCT
- Discuss the similarity between redwood or tanoak basal area and vegetation fuel loading
- Discuss the potential over-estimation of vegetation fuel loading
4.1 Regeneration
4.1.1 Redwood and tanoak
Basal area was selected as the metric for quantifying redwood and tanoak abundance because prolific sprout regeneration in both species was undergoing self-thinning. In this context, basal area provided a more informative measure than stem density, as it better captured treatment responses by reflecting the relative contribution of established stems rather than the transient abundance of sprouts.
We found redwood growth—in terms of average total basal area of sprouts 10 years after harvest treatment—responded strongly and positively to increasing levels of openness consistent with previous findings that redwood sprout growth is highly sensitive to light availability and overstory density (Berrill et al., 2018; R. Muma et al., 2022; O’Hara et al., 2007; O’Hara & Berrill, 2010). In contrast, the tanoak response was comparatively modest reflecting its typical habit of forming a sub-canopy layer in these systems (O’Hara et al., 2017).
Although none of the treatment differences for tanoak were statistically supported, the largest difference was between the GS and HA treatments, suggesting that the HA treatment minimized tanoak sprout growth.
Shade tolerance theory predicts that at a certain overstory density, tanoak abundance should exceed that of redwood. This crossover point appears to have been approached in the HA and HD treatments where basal area for these two species were very similar.
It is important to point out here that the HA and HD “high-density” treatments in this experiment only targeted a residual overstory relative density of 20% (that is 20% of assumed total carrying capacity of the site), and this was selected as an upper limit to support the objective of maintaining conifer growth (Berrill & O’Hara, 2009). It seems likely that at higher overstory densities we might see tanoak abundance, in terms of total basal area, exceed that of redwood in the understory.
Our use of basal area as a response did not distinguish between number of stems and size of stems. There have been numerous metrics used in attempts to assess shade tolerance (Forrester et al., 2014). Our use of basal area (which conflates growth and survival of sprouts) is justified given the assumption that these are expected to be correlated in forests that don’t undergo a long harsh winter (Lin et al., 2002).
Little is understood about tanoak sprout response under shade (Waring & O’Hara, 2008; Wilkinson et al., 1997). This study contributes to that knowledge by describing a range of tanoak sprout development responses with varying overstory density. Our results suggest that tanoak sprout basal area is reduced by about half when moving from open overstory conditions to 20% relative density.
Average redwood sprout basal area 10 years after treatment was much more dramatically affected by overstory density and each step increase in overstory density (GS -> LD -> HA) saw average redwood sprout basal area reduced by around half. Aggregated retention resulted in about 1/3 greater basal area than in the dispersed retention but the p-values for all comparisons were very high (p >= 0.2).
In all treatments, the uncertainty in the average redwood basal area is much greater than that of tanoak. Given that we modeled the total basal area of all species in each vegetation plot, this suggests that variability in redwood basal area was much greater than that of tanoak across macro plots and/or sites. The greater relative uncertainty we observed in redwood compared to tanoak may have to do with redwoods greater sensitivity to light conditions. We selected sites with both north and south facing slopes and this may have led to relatively larger differences for redwoods growth response compared to tanoak. Additional uncertainty may be due to variable light levels within macro plots.
It is interesting to note that basal area comparisons between redwood and tanoak resulted in steadily decreasing differences between these two species as overstory density increases. This is in contrast to findings from this and other studies when the focus is instead on sprout height growth. This perspective of growth response highlights an important question for these forests: what is the role of below-ground competition in determining the relative growth of redwood and tanoak sprouts? This question has been raised in other studies for redwoods (Oliver et al., 1994) and has been found to play play an important role in moderating growth beyond what was expected from shade-tolerance alone, particularly with regard to soil quality (Forrester et al., 2014), or size of cut stump as a proxy for size or resprouting root system (Berrill et al., 2018).
4.1.2 Douglas-fir counts
Douglas-fir abundance was modeled using stem counts because the species does not regenerate through sprouting and thus did not exhibit the self-thinning found in the sprouting species. Moreover, stems were relatively few in number and consisted largely of small seedlings, making stem density a more informative metric than basal area for comparing responses across treatments.
Douglas-fir is a minor species of special interest due to it’s commercial value and ability to compete with redwoods in the first 100 years of stand development (Wensel & Krumland, 1986). We didn’t detect statistically supported differences between treatments in terms of basal area or stem counts. While our observed recruitment density of Douglas-fir seedlings may be sufficient for the eventual production of a viable cohort in a multi-aged system (Schütz and Röhnisch, 2003, as cited in Schütz & Pommerening, 2013), only the LD and GS treatments are likely to maintain overstory densities conducive to Douglas-fir seedling development (Miller & Emmingham, 2001; Schütz & Pommerening, 2013) given the overstory basal areas recorded at the initiation of the experiment (R. Muma et al., 2022). Rapidly developing redwood and tanoak sprouts will likely further comprise Douglas-fir seedlings competitive ability.
4.1.3 Other species
Our sampling included relatively few minor species and differences across treatments were small. Though none of the treatment differences were statistically supported, the largest difference occurred between the LD and HA treatment where the latter had less basal area of minor species. While this could be due to low numbers of occurrences of minor species and high variability within treatments, this uniqueness of the HA treatment is found in other places throughout our experiment results: such as for tanoak sprout basal area where the greatest treatment difference was between the GS and HA treatments and in the low vegetation response found in the HA treatment for the pre-PCT fuels data. Potential differences in minor species’ abundances is likely only relevant for grand fir, due to its shade tolerance (Webb et al., 2012). Red alder and western hemlock were observed during an earlier iteration of this experiment, but were not detected with our current experimental design (R. T. Muma, 2019).
4.2 Sprout height
We found that for both redwood and tanoak, height increment and overall height at year 10 were greatest in the Group Selection (GS) treatment, followed by Low-Density Dispersed (LD), High-Density Aggregated (HA), and High-Density Dispersed (HD) treatments. This pattern indicates that greater reductions in overstory density lead to more rapid sprout growth.
This strong positive relationship between reduced overstory density and sprout height is a consistent and well-documented finding across various species and forest types in the literature: Understory light availability, which is inversely related to overstory density, is a primary driver of stump sprout growth in redwood-mixed-conifer systems (Berrill et al., 2018; Berrill et al., 2021), and in other systems as well (Atwood et al., 2009; Gardiner & Helmig, 1997; Keyser & Zarnoch, 2014; Knapp et al., 2017).
We found that redwood grew faster and achieved greater heights than tanoak across all treatments and growth periods. For example, redwood height increment was 0.42 m/yr greater than tanoak in the first period and 0.33 m/yr greater in the second period. Also, redwood height growth response was much stronger in more open conditions, while tanoak’s was comparatively modest.
The increased overstory reduction in this experiment successfully maintained redwood sprout growth, compared to a previous study where overstory reduction was not sufficient to maintain redwood sprout growth (O’Hara & Berrill, 2010). Despite substantial reductions in redwood basal area with our high-overstory-retention treatments, especially compared to tanoak, redwood height growth was still dominant at about 0.6 m yr-1. While the aggregated high-density treatment saw marginally higher sprout height development than the HD treatment, the difference was not large or statistically significant. The initial harvest in both high-density treatments in this experiment resulted in about 39 m2 ha-1. This suggests that the level of overstory harvest prescribed by a previous growth modeling study (32-38 m2 ha-1, Berrill & O’Hara, 2009) may be an upper bound in these systems if redwood sprout growth is to be maintained until the next partial harvest.
4.3 Fuel loading
4.3.1 Pre-PCT
Prior to PCT scheduled Fuel loading were comparable to other studies in redwood systems. Our treatment level averages for duff and litter ranged from 40 to 55 Mg ha-1 in plots that had post-harvest basal areas (in 2012) of 0 to 40 m2 ha-1. This was comparable to total duff and litter loading found in 120-year-old redwood stands (range: 29 to 55 Mg ha-1) as well as old-growth stands (average: 50 Mg ha-1) and our raw average duff and litter depth of 6.2 cm (data not shown) was similar to that found in a range of redwood stand structures (Finney & Martin, 1993a; Glebocki, 2015; Stuart, 1985). We opted to combine duff and litter into a single metric following previous studies that found similar bulk densities for duff and litter and that separating the two resulted in little difference from and average bulk density, given the wide variability found in duff and litter depths Finney & Martin (1993a).
Our average total fine woody fuel loading, including one-, ten-, and one hundred-hour-fuels (15 Mg ha-1) was within the range found in 120-year-old stands (9 to 20 Mg ha-1, Finney & Martin (1993b)), double that found in very young, mixed stands (7 Mg ha-1, Glebocki (2015)), and 3 Mg ha-1 higher than that found in old growth stands (12 Mg ha-1, Stuart (1985)).
In the following I’ll compare the average fuel loadings found in this study to those found in the “old growth” and “very young, mixed” stands referenced above.
Our one-hr fuel loading which, which averaged about 1 Mg ha-1 was similar to that found in the old-growth stand, but about half that found in the very young, mixed stand, possibly due to differing stand structures and species compositions. That young forest was dominated by young Douglas-fir and would be expected to have different forest floor characteristics and this notion is supported by our finding of similar 1-hr fuel loading as in the old growth study. Another source of possible error is differences in sampling method for 1-hr fuels. We established a cutoff where redwood particles smaller than roughly 2 mm were considered as litter, rather than 1-hr fuels. If we had counted every redwood leaf-spray, regardless of size, we would have likely found higher 1-hr fuel loads.
Our average 10-hr fuel loading was higher than in the very young, mixed stand (3.4 vs. 2 Mg ha-1) but similar to the old growth stand. It might be reasonable here to suspect that the smaller trees in the very young, mixed stand did not supply as much 10-hr fuels because of the sizes of their branches, whereas branch shedding patterns in the 10-hr time lags class were more similar between our stands and the old growth ones.
Our average 100-hr fuel loading of about 11 Mg ha-1 was higher than in both the old growth stands (5 Mg ha-1) and the very young, mixed stands (3 Mg ha-1). This may be the result of residual fuels left over from the harvest treatment that initiated our experiment which would have been absent from the other sites referenced here. In the very young, mixed stand for instance, 100-hr fuels jumped to 10 Mg ha-1 immediately following a thinning treatment.
Our average 1,000-hr fuel loading of 42 Mg ha-1 was somewhat lower than found in old growth stands (63 Mg ha-1) and the very young, mixed stands (54 Mg ha-1). These estimates are accompanied by relatively high standard errors, but it would not be surprising that State Park forests, with little to no recent harvest activity would have a larger amount of large downed logs then an actively managed forest such as ours.
The similarity in average fuel loading between our Pre-PCT stands and two studies conducted in structurally distinct redwood forests aligns with our finding of few statistically significant differences in fuel loading among treatments. This result is common among other studies and is expected due to the highly variable nature of forest fuels (Collins et al., 2016; Keane et al., 2001, 2012).
TODO: discuss observed anomalies with HA treatment in general
In all cases the largest observed differences involved the HA treatment. For instance, for Duff & Litter was lowest in the HA treatment and highest in the HD treatment (p = 0.10) and this trend was paralleled by a similar trend in vegetation fuel loading as well as in tanoak basal area.
Vegetation fuel loading was lowest in the HA treatment and the greatest difference was between the GS and HA treatments. Lower vegetation fuel loading in the HA treatment is reflected in the tanoak basal area results for the HA treatment, which were collected separately
In a study across California’s Sierra Nevada mountain range, fine fuel loading (including 1-, 10-, and 100-hr fuels) were found to be correlated with overstory canopy cover and proportion of shade tolerant species and associated these conditions, which represent recent shifts in forest composition resulting from fire suppression, with an elevated risk of high severity fire (Collins et al., 2016). The differences we observed in duff & litter, and 1-hr fuel loads could be a result of canopy cover and understory species composition directly, but it is likely also a result of the microclimatic differences resulting from these different stand structures. Fine fuel loads are a function of both recruitment and decomposition and the latter is a function of moisture and temperature (Keane, 2008; Pillers & Stuart, 1993). Forest floor conditions in the HA treatment may have been somewhat more moist than the more open treatments due to sheltering but warmer than the HD treatment due to greater light infiltration (e.g., Kane (2021)). While the differences in duff & litter loading we observed are not likely to have a significant impact on fire behavior, small differences in forest floor moisture can have significant implications for prescribed burning operations.
For both 1-hr and vegetation fuel loading, the HA treatment was most similar to the HD treatment. In both treatments it appears that a greater density of overstory trees would result in more twigs and less light available for growth of vegetation at the surface. The high density treatments had less tanoak BA and it is likely that tanoak was largely responsible for the vegetative fuel loading differences. Vegetation fuel loading was not recorded by species but redwood and tanoak sprouts comprised the vast majority of vegetation fuels.
Consistently low foliar moisture content levels have been recorded for tanaok (Kuljian & Varner, 2010) and the foliage has been compared to that of other sclerophyllous species such as those found in chaparral ecosystems which are characterized by intense, stand replacing fires suggesting high flammability, particularly with low canopy base heights (Fryer, 2008; McDonald, 1981). And tanoak leaf litter is among the most flammable of western hardwoods and has been compared to that of fire adapted conifers (Varner et al., 2017). To our knowledge, no studies have examined the flammability of live tanoak foliage or tanoak sprouts.
Our understanding of the conditions under which crown fires initiate and spread are not well understood for several reasons and largely because of the difficulty in observing and studying large fires (Finney et al., 2021; Xanthopoulos & Athanasiou, 2020). Less still is known about crown fire behavior in this unique and transitional fuel type (mixed conifer/tanoak sprouts). Because of the relatively low bulk density of canopy fuels, smaller differences in fuel load equate to larger differences in total fuel volume and resulting fire behavior, as well as implications for aerial fuel continuity. These factors warrant more study into the fire behavior of this unique fuel type, including its requirements for combustion, fuel moisture, and more precise characterization of it’s mass and bulk density.
It was surprising that the modeled vegetation fuel loading differences were not more clearly differentiated because our field crew’s experience was that the GS and LD treatments were usually much more difficult to work in and travel through due to greater understory density.
In this study, live and dead vegetation components were combined because the focus was on total fuel load rather than fuel moisture content. This choice allowed for a clearer representation of the overall quantity of combustible material. Additionally, herbaceous fuel loading was generally very low, which supported the decision to merge herbaceous and woody loads at the transect level. By simplifying in this way, the analysis emphasized patterns of total vegetation loading while limiting unnecessary complexity in the analysis.
4.3.2 Post-PCT
Duff and litter are not reported for post-pct stands for two reasons. First, these are not expected to have changed significantly given the relatively short timespan before and after PCT. Second, our sampling protocol did not make it clear how to quantify the loading for leaves attached to recently cut branches, especially given that these “suspended litter” particles were in a state of active transition to the ground, as they dry, abscise, and sift through the coarser woody fuels as they make their way to the ground. This class may deserve more attention because of it’s potentially dynamic relationship with the timing of prescribed fires: as suspended particles settle and begin to decompose they’re bulk density changes and bulk density is an important determinant of fire behavior.
Vegetative growth following thinning and harvest is an important consideration for fire informed management. In our experiment, greater overstory retention resulted in lower understory size and density. Therefore, 10 years of growth (followed by PCT) led to some increase in average fine woody fuel loading, particularly in the dense understories of the GS and LD treatments. One-hr fuels increased most in the GS and LD treatments but there was also an unexpected increase in 1-hr fuels in the HD treatment compared to the HA treatment which had a negligible increase. Ten-hr fuels increased most in the GS and somewhat in the LD treatment, reflecting the more rapid growth seen in these treatments compared to the others.
Pre-commercial thinning resulted in little change in 1,000-hr fuels, as thinned sapling stems were mostly under three inches, although a potential increase observed in the HD treatment may have been a result of some previously retained canopy trees being cut (12 inches DBH was the prescribed upper limit for thinning). The model selection process resulted in slightly different models after pre-commercial thinning, compared to before thinning for a given fuel class. The model selection process was guided by balancing parsimony, AIC, and the production of well distributed residuals. As predicted fuel load variability increased along with predicted fuel loads following PCT, model selection tended towards more complex models that accounted for this variability–particularly for 1- and 100-hr fuels.
TODO: I’m here
It can be seen from this analysis, resulting surfaces fuel conditions are expected to be the result of a complex interaction of productivity and decomposition rate. Whether any of the differences we observed result in a significant difference in prescribed or wild fire will need to be tested by those disturbances. It is possible that the elevated 10-hr loads in the GS and to a lesser degree, the LD treatments may represent greater fuel continuity which could support prescribed fire operations. Likewise, the HA treatment’s lower fuel loading across classes could signify patchy fuel distribution. Surface fuel moisture and wind speed are two variables which we did not study but which have a strong influence on fire behavior and are also affected by overstory density and arrangement. Given the relatively modest difference we observed in fuel loads, it seems likely that those unmeasured factors would have a greater bearing on fire outcomes.
5 Limitations and future work
Throughout this multi-faceted analysis there were several limitations that arose.
Basal area comparisons among treatments and with other species within a treatment for tanoak were confounded by the fact that tanoak produced an abundance of small diameter sprouts, especially in the HA and HD treatments treatment (383 in HA vs 496 in HD, data not shown). Stems smaller than 2.54 cm DBH were tallied. To facilitate basal area analysis, I assigned each of these the possibly inaccurate midpoint diameter of 1.27 cm. Thus, the basal area of these bushier tanaok in the HD and HA treatments may be slightly artificially elevated. On the other hand, the general trend of treatment effect on tanaok basal area response was mirrored in the vegetation fuel loading data which was collected separately, lending support to the idea that the HA treatment may have minimized tanoak sprout growth.
In the analysis of sprout basal area, aspect would be an important consideration for sprouts growth response. Reducing our uncertainty around redwood basal area may require better incorporating the effect of aspect into our model. Also regarding aspect, it seems likely that the effect of site on sprout growth was bi-modal (due to our combination of predominately north- and south-facing site aspects) and thus the regularization imposed by our random effects for sites (assumed to be normally distributed) was likely inappropriate.
Our method of calculating vegetation fuel loading (estimated percent cover times estimated height within an estimated cylinder) might have been overly imprecise. Additionally, given the importance of understory light to the processes of surface fuel decomposition as well as for understory growth, including aspect might have led to better explanatory power in our models. Our plots were established over a range of aspects, but it is plausible that differences in aspect outweighed differences between treatments. We had hoped that the inclusion of site and treatment as nested random effects would have captured site and block level differences in aspect. Additionally, aspect is largely a proxy for insolation, and this may have also varied with shade conditions (very large trees or road cuts) outside of the plot.
The simple method we used probably over-estimated vegetation fuel loading, evidenced by the fact that the large fuel loading differences measured for the vegetation class are not reflected in the increase in fine fuels following PCT. Although the contribution of slash foliage is not accounted for, it appears our method may have over predicted vegetation loading by around a factor of three under the assumption that the majority of the difference in vegetation fuels should be captured by the sum of the differences in fine fuels.
An important consideration for redwood sprouts in this experiment is that after thinning, a fraction of existing sprouts will be retained. The retained sprouts will generally be the largest and most vigorous. An interesting question that is overlooked by our focus on the total basal area of sprout regeneration is if any treatment produced larger individuals that would eventually be selected for retention.