Simulated Fire Behavior and Fine-Scale Forest Structure Following Conifer Removal in Aspen-Conifer Forests in the Lake Tahoe Basin, USA

: Quaking aspen is found in western forests of the United States and is currently at risk of loss due to conifer competition at within-stand scales. Wildﬁres in these forests are impactful owing to conifer inﬁlling during prolonged ﬁre suppression post-Euro-American settlement. Here, restoration cuttings seek to impact wildﬁre behavior and aspen growing conditions. In this study, we explored how actual and hypothetical cuttings with a range of conifer removal intensity altered surface fuel and overstory structure at stand and ﬁne scales. We then simulated wildﬁres, examining ﬁre behavior and e ﬀ ects on post-ﬁre forest structures around aspen trees. We found that conifer removal constrained by lower upper diameter limits ( < 56 cm) had marginal e ﬀ ects on surface fuel and overstory structure, likely failing to enhance resource conditions su ﬃ ciently to sustain aspen. Increasing the diameter limit also led to a higher likelihood of ﬁre spread and a higher rate of spread, owing to greater within-canopy wind speed, though crown ﬁre activity decreased. Our simulations suggest heavier treatments could facilitate reintroduction of ﬁre while also dampening the e ﬀ ects of wildﬁres on forest structure. Cutting speciﬁcations that relax diameter limits and remove a substantial portion of conifer overstory could better promote aspen restoration and mitigate ﬁre hazard.


Introduction
Land management activities for over a century have led to significant changes in forest structure and ecosystem processes in mixed conifer-quaking aspen (Populus tremuloides Michx.) forests where wildland fire once regulated forest developmental pathways [1][2][3][4]. In these forests, low-and mixed-severity fires produced openings sufficiently large to modify microclimates, thus yielding fine-scale diversity of growing environments and regeneration niches [3,5,6]. However, lack of fires through suppression, as well as droughtiness, and livestock herbivory have reduced aspen vigor and recruitment, leading to replacement by conifers [1,[7][8][9][10].
Loss of aspen within mixed forests is particularly acute in regions like the Lake Tahoe Basin of the western US, where aspen trees are typically found in relatively small (<1.5 ha) patches. Despite covering less than 2% of the Lake Tahoe Basin [7], forest managers and scientists have identified these mixed forest plots previously reported by Berrill et al. [14]. Forest managers decided these three plots were most representative of the landscape managed by the LTWRP.
These plots were rectangular, measuring 142.86 m × 70 m for BP2 and 200 m × 50 m for WA38 and CV05. All aspen with a diameter at breast height (dbh) above 10 cm and conifer trees with a dbh above 20 cm were mapped and had the following information recorded: species, dbh, tree height, and crown base height [18]. Coniferous species on these sites included lodgepole and Jeffrey pine (Pinus contorta Douglas ex Loudon var. murrayana (Balf.) Engelm.; P. jeffreyi Grev. & Balf.) and white and red fir (Abies concolor (Gord. & Glend.) Lindl. ex Hildebr.; A. magnifica A. Murr.).
We simulated three thin-from-below cuttings to assess the effects of varying conifer removal, including a light, medium, and heavy cutting, removing conifers up to 36 cm, 56 cm, and 76 cm dbh, respectively. We chose these scenarios in consultation with local land managers to span a range of either implemented or desirable cutting intensities in the Lake Tahoe Basin. In addition to these hypothetical scenarios, we evaluated the impacts of actual treatments that occurred on two of the sites (WA38 and CV05) based on re-inventoried post-treatment field data. These two observed treatments emphasized removing conifers to benefit aspen and lowering fire behavior and effects. The maximum size limit for conifer removal for the observed cutting treatments was 35 cm and 30 cm on sites WA38 and CV05, respectively.

Simulated Fire Behavior and Effects on Conifers
We simulated fire behavior with the Wildland urban-interface Fire Dynamics Simulator (WFDS) version 9977 [23]. WFDS is a physics-based computational model of wildland fire, which extends the Fire Dynamics Simulator (FDS) developed by the National Institute of Standards and Technology [24] to include vegetative fuel. WFDS simulates the spatial and temporal dynamics of wildland fire through the explicit representation of fire-atmosphere-fuel dynamics [25]. This is achieved by implementing computational fluid dynamics methods to solve the conservation of mass, momentum, and energy equations along with models for radiative heat transfer, thermal degradation of fuel, and gas-phase combustion. Wildland fuels are described in the model in three dimensions based on their geometry (e.g., crown base height, width, length, and shape) and their bulk properties (e.g., bulk density, and fuel moisture). Additional references provide model details and verification and validation studies [24,[26][27][28][29][30][31][32][33][34].
We simulated potential fire behavior and effects for each site and cutting intensity under three wind speeds, defined by different inlet open wind speeds. In all simulations, we used an overall domain size measuring 1000 m × 320 m × 100 m in the x, y, and z dimensions, respectively. Similar to other studies using WFDS [35], the domain design ( Figure 1) contained four critical parts: (1) a wind development zone located at x = [0, 400] m; (2) an ignition burner at x = [400, 430] m, y = [100, 220] m; (3) our area of interest, defined by the size of the study sites and beginning at x = 500 m; and (4) the outflow zone extending to the outlet at x = 1000 m. The domain was discretized into 2 m 3 cells, except the volume above the area of interest, which was discretized at a 1 m 3 resolution. We parameterized the inlet wind field using a standard logarithmic wind profile with the open wind speed (z = 40 m) as either 6, 12, or 18 m/s representing our low, moderate, and high wind speed scenarios, respectively. The side and top boundaries were simulated as no flux, no slip surfaces, and the outflow boundary was open, allowing free atmospheric exchange into or out of the domain. The bottom boundary was flat (i.e., no slope).
For each simulation, we populated the area of interest with the tree list of a given study site's cutting scenario. We represented conifer and aspen trees as right circular cones and cylinders, respectively, defined by their location, tree height, crown base height, and crown width. Crown width was estimated using dbh-based, species-specific allometry [36,37]. We populated the rest of each simulation's domain with additional trees to ensure the wind fields in our area of interest were representative of analogous forests. Those additional trees were randomly distributed using a log-gaussian Cox point process model fit on the observed aspen and conifer trees within the area of interest [38]. A point process model ensured that the locations of randomly placed trees emulated the tree spatial pattern of measured trees. For each simulation, we populated the area of interest with the tree list of a given study site's cutting scenario. We represented conifer and aspen trees as right circular cones and cylinders, respectively, defined by their location, tree height, crown base height, and crown width. Crown width was estimated using dbh-based, species-specific allometry [36,37]. We populated the rest of each simulation's domain with additional trees to ensure the wind fields in our area of interest were representative of analogous forests. Those additional trees were randomly distributed using a log-gaussian Cox point process model fit on the observed aspen and conifer trees within the area of interest [38]. A point process model ensured that the locations of randomly placed trees emulated the tree spatial pattern of measured trees.
We simulated surface fuels as a heterogeneous mixture of fuelbeds based on the local density of the overstory species composition. These fuelbeds were developed locally for use in the Lake Tahoe Basin [39]. To this end, we first estimated the density of trees within three species categories, aspen (quaking aspen), fir (red fir), or pine (lodgepole pine and white fir), in each 1 m 2 cell using a spatial smoothing algorithm that includes a Gaussian kernel with a standard deviation of 1 m 2 and an edge correction scheme [38]. A fuelbed type exists for Jeffrey pine [39], but no cell was dominated by Jeffrey pine. Next, we categorized each 1 m 2 cell according to whichever category's smoothed density was greatest. Finally, we related each of these categories to a customized fuelbed (Table 1). We simulated surface fuels as a heterogeneous mixture of fuelbeds based on the local density of the overstory species composition. These fuelbeds were developed locally for use in the Lake Tahoe Basin [39]. To this end, we first estimated the density of trees within three species categories, aspen (quaking aspen), fir (red fir), or pine (lodgepole pine and white fir), in each 1 m 2 cell using a spatial smoothing algorithm that includes a Gaussian kernel with a standard deviation of 1 m 2 and an edge correction scheme [38]. A fuelbed type exists for Jeffrey pine [39], but no cell was dominated by Jeffrey pine. Next, we categorized each 1 m 2 cell according to whichever category's smoothed density was greatest. Finally, we related each of these categories to a customized fuelbed (Table 1). We measured mid-flame wind speed, mean rate of fire spread, and total canopy consumption within the area of interest of each simulation. Omitting the simulations on the observed treatments, we assessed the impacts of cutting intensity, wind speed, and their interaction on these three measurements using mixed-effects ANOVAs; sites were included as a random variable. For all tests in this study, α = 0.05.

Effects of Conifer Removal and Fire on Stand Structure
We used individual tree crown consumption as a proxy for mortality, assuming conifers with any amount of crown consumption were killed. Though many factors, including necrosis of tissues, deformation of xylem, and post-fire plant stressors, affect the likelihood of fire-induced tree mortality [40], several studies have suggested that crown consumption is a powerful predictor of conifer mortality [40][41][42].
We tracked spatially averaged stand metrics and local aspen-centric metrics to investigate the effects of cuttings and subsequent fires on forest structure. First, we examined changes in basal area, trees per hectare, and canopy cover at the stand scale. Canopy cover was calculated as the percent of area occupied by the sum of non-overlapping projected crown areas. Next, to assess local effects of conifer removal on aspen, we calculated the proportion of aspen that have a conifer present within 1.5 tree lengths and the stand density index (SDI) of those conifers. We used mixed-effects ANOVAs, where the dependent variable was the change from pre-fire to post-fire for each of the forest structure metrics. Fixed variables included cutting intensity, wind speed, and their interaction; study site was a random variable.

General Effects of Cutting and Fire
All three sites had a similar basal area before cutting (~50 m 2 ha −1 ) but differed in their density (341 to 555 trees ha −1 ) and the proportion and size of conifers present (Table 2). While the light and observed cuttings decreased tree density and increased quadratic mean diameter, other measures of forest structure were not substantially reduced until cutting intensity approached the medium and high levels of conifer removal (Table 2).
In particular, the simulated light cuttings reduced conifer density by 54% and conifer basal area by 24%, averaged across sites. Total canopy cover remained relatively high (51-63%) with conifer cover greater than 25% on all sites. Light cuttings resulted in a marginal reduction in the proportion of aspen with a conifer present within 1.5 tree lengths (1-2%) and reduced conifer SDI near aspen by 23-40%. The observed cutting treatments tended to mirror the light cutting scenarios, although the observed cuttings removed only 31% and 51% of conifers in sites CV05 and WA38, respectively.
In contrast to the light cuttings, the medium cuttings reduced conifer density by 87% (34% more than light cuttings) and conifer basal area by 61% (37% more than light cuttings) averaged across sites. Total canopy cover was reduced to 42-53%, of which conifer cover was 18-27%. Conifers within 1.5 tree lengths were present near 75-93% of aspen across the sites. The SDI of these proximate conifers ranged from 49 to 26 on average, just one-third of pre-treatment local SDI.
Heavy cuttings markedly reduced conifer abundance. After heavy cutting, only 5 to 19 conifers ha −1 remained, amounting to only 4% of the original conifers, or 17% of initial conifer basal area, averaged across sites. The average residual canopy cover was around 55% (average conifer canopy was 16%). Conifers were proximate to only half or slightly greater of the aspen, resulting in a local SDI reduced by 60-80% of pre-treatment values.  Table 2. Summary of forest structure and local stocking proximate to aspen pre-cutting; after simulated light (removal of conifers <36 cm dbh), medium (removal of conifers <56 cm dbh), and heavy (removal of conifers <76 cm dbh) cuttings; and in the observed post-treatment cutting. QMD and SDI are quadratic mean diameter and stand density index, respectively. Though all scenarios retained a heterogeneous mixture of fuelbed types, cuttings had a significant impact on surface fuelbed composition ( Figure 2). Thirty-two, 45, and 49 percent of the surface fuels were classified as aspen fuelbeds for WA38, BP2, and CV05, respectively, before any treatment. Aspen fuelbeds became the predominant fuel type, as the cutting intensity increased, primarily at the expense of fir ( Figure 2). The light and observed cutting scenarios had relatively minor effects on the proportion and patch size of aspen surface fuels. In comparison, the medium and heavy cutting scenarios increased the proportion of aspen fuelbeds by 70 to 90 percent. As fuel loads varied by fuelbed type, cuttings concomitantly reduced the average fuel loads across the study sites (Table 3); light, medium, heavy, and the observed cuttings reduced the average fuel loads by 17%, 33%, 39%, and 11%, respectively. Fire 2020, 3, x FOR PEER REVIEW 6 of 16 canopy cover was reduced to 42-53%, of which conifer cover was 18-27%. Conifers within 1.5 tree lengths were present near 75-93% of aspen across the sites. The SDI of these proximate conifers ranged from 49 to 26 on average, just one-third of pre-treatment local SDI. Heavy cuttings markedly reduced conifer abundance. After heavy cutting, only 5 to 19 conifers ha −1 remained, amounting to only 4% of the original conifers, or 17% of initial conifer basal area, averaged across sites. The average residual canopy cover was around 55% (average conifer canopy was 16%). Conifers were proximate to only half or slightly greater of the aspen, resulting in a local SDI reduced by 60-80% of pre-treatment values.

Site/Cutting
Though all scenarios retained a heterogeneous mixture of fuelbed types, cuttings had a significant impact on surface fuelbed composition (Figure 2). Thirty-two, 45, and 49 percent of the surface fuels were classified as aspen fuelbeds for WA38, BP2, and CV05, respectively, before any treatment. Aspen fuelbeds became the predominant fuel type, as the cutting intensity increased, primarily at the expense of fir ( Figure  2). The light and observed cutting scenarios had relatively minor effects on the proportion and patch size of aspen surface fuels. In comparison, the medium and heavy cutting scenarios increased the proportion of aspen fuelbeds by 70 to 90 percent. As fuel loads varied by fuelbed type, cuttings concomitantly reduced the average fuel loads across the study sites (Table 3); light, medium, heavy, and the observed cuttings reduced the average fuel loads by 17%, 33%, 39%, and 11%, respectively.   Fire behavior varied among the cutting and wind speed scenarios. Fire spread across our plots was particularly conditional on within-stand wind speeds. Before cutting, or following cuttings of low intensity, fire spread in sites BP2 and CV05 either ceased or was not uniform (Figure 3). Higher wind speeds facilitated more comprehensive fire spread in these instances. The capacity for a fire to burn throughout the plots under more intense cuttings, despite reductions in fuel load is partly explained by increases in the mid-flame wind speed (Figure 4a). Further reductions in conifer stocking led to higher mid-flame wind speeds as the canopy opened (F 1,30 = 83.10, p < 0.001). There was also an interaction between cutting intensity and mid-flame wind speed; when we simulated fires with higher open wind speeds, greater cutting intensities led to even greater mid-flame wind speeds (F 1,30 = 18.71, p < 0.001). Accordingly, rate of fire spread increased with greater intensity of cutting (F 1,30 = 45.68, p < 0.001), and greater wind speed (F 1,30 = 79.59, p < 0.001); however, there was no clear interaction between intensity of cutting and wind speed (F 1,30 = 2.64, p = 0.115). Visual inspection of the rate of fire spread shows a tendency for spread rate to slightly diminish under heavy cutting relative to medium cutting. This indicates a crossover point regarding a trade-off between higher mid-flame wind speeds and lower surface fuel load associated with aspen ( Figure 4b).
Despite increased mid-flame wind speeds associated with cutting and rate of fire spread, we found the amount of canopy consumption decreased greatly with cutting intensity (F 1,30 = 58.75, p < 0.001), particularly between light and medium cuttings (Figure 4c). This occurred despite slight increases of canopy consumption with greater wind speed (F 1,30 = 4.2, p = 0.049). Though the effect size was marginal, visual inspection of Figure 4c suggests that greater cutting intensity was more effective at reducing canopy consumption when wind speed increased (F 1,30 = 3.34, p = 0.077). Last, we note that fire behavior within the observed cuttings was generally similar to either the light cuttings or pre-cutting scenarios.
The effect of fire on conifer mortality varied across cutting and wind speed scenarios ( Figure 5). Fires reduced conifer canopy cover least when study sites had been cut more heavily (F 1,30 = 94.56, p < 0.001) or under low wind speeds (F 1,30 = 4.00, p = 0.050). Further, the effect of cutting intensity depended on wind speed (F 1,30 = 4.46, p = 0.040); when cuttings were lighter, or in the pre-cutting scenario, canopy consumption increased with greater wind speeds (Figure 5a). When cutting intensity was greater, the magnitude of open wind speed had little effect.
Though fires did reduce conifer cover, particularly following no or light cuttings, conifers were still near quaking aspen. Figure 5b shows that fires slightly reduced the proportion of aspen with local conifer presence; this reduction did not vary with cutting intensity (F 1,30 = 2.07, p = 0.160) nor did cutting intensity interact with wind speed (F 1,30 = 0.92, p = 0.344). Only the effect of wind speed was impactful (F 1,30 = 6.53, p = 0.016) with greater wind speed leading to decreases in post-fire conifer proximate to aspen.
Cutting intensity did, however, influence the local stocking around aspen post-fire ( Figure 5c). Fires decreased local SDI most under lighter cuttings (F 1,30 = 7.84, p = 0.009) and greater wind speeds (F 1,30 = 6.91, p = 0.013). In addition, local SDI was most impacted by fire in the cases of no or light cutting and when wind speed was greater (F 1,30 = 9.23, p = 0.005). For all these metrics, the effect of fire after observed cuttings was similar to light cuttings ( Figure 5).   The effect of fire on conifer mortality varied across cutting and wind speed scenarios ( Figure 5). Fires reduced conifer canopy cover least when study sites had been cut more heavily (F1,30 = 94.56, p < 0.001) or under low wind speeds (F1,30 = 4.00, p = 0.050). Further, the effect of cutting intensity depended on wind speed (F1,30 = 4.46, p = 0.040); when cuttings were lighter, or in the pre-cutting scenario, canopy consumption increased with greater wind speeds (Figure 5a). When cutting intensity was greater, the magnitude of open wind speed had little effect.
Though fires did reduce conifer cover, particularly following no or light cuttings, conifers were still near quaking aspen. Figure 5b shows that fires slightly reduced the proportion of aspen with local conifer presence; this reduction did not vary with cutting intensity (F1,30 = 2.07, p = 0.160) nor did cutting intensity interact with wind speed (F1,30 = 0.92, p = 0.344). Only the effect of wind speed was impactful (F1,30 = 6.53, p = 0.016) with greater wind speed leading to decreases in post-fire conifer proximate to aspen. Cutting intensity did, however, influence the local stocking around aspen post-fire ( Figure 5c). Fires decreased local SDI most under lighter cuttings (F1,30 = 7.84, p = 0.009) and greater wind speeds (F1,30 = 6.91, p = 0.013). In addition, local SDI was most impacted by fire in the cases of no or light cutting and when wind speed was greater (F1,30 = 9.23, p = 0.005). For all these metrics, the effect of fire after observed cuttings was similar to light cuttings ( Figure 5).

Discussion
Our results demonstrate how a significant removal of moderate to large conifer trees may promote restoration of aspen by alleviating conifer crowding around aspen and altering fuel and fire behavior. Even though the actual implemented treatments removed one-third to one-half of conifers, substantial changes to stand structure and simulated fires did not manifest until we simulated removal of over 80% of conifers, leaving behind less than c. 50 conifers ha −1 or 9 m 2 ha −1 basal area. Furthermore, lighter treatments did not meet several rules of thumb for aspen restoration, namely, reducing conifer cover to under 25% and removing conifer trees within 1.5 tree lengths of aspen [7,14,17]. Our results corroborate findings made by others in the Lake Tahoe Basin, which suggest that aspen restoration treatments may better meet desired ecological processes and stand dynamics by raising diameter limits [9,13,18,19].

Discussion
Our results demonstrate how a significant removal of moderate to large conifer trees may promote restoration of aspen by alleviating conifer crowding around aspen and altering fuel and fire behavior. Even though the actual implemented treatments removed one-third to one-half of conifers, substantial changes to stand structure and simulated fires did not manifest until we simulated removal of over 80% of conifers, leaving behind less than c. 50 conifers ha −1 or 9 m 2 ha −1 basal area. Furthermore, lighter treatments did not meet several rules of thumb for aspen restoration, namely, reducing conifer cover to under 25% and removing conifer trees within 1.5 tree lengths of aspen [7,14,17]. Our results corroborate findings made by others in the Lake Tahoe Basin, which suggest that aspen restoration treatments may better meet desired ecological processes and stand dynamics by raising diameter limits [9,13,18,19].
Fires may be complementary to aspen restoration cuttings by reducing conifer stocking, but the efficacy of fire depends on cutting intensity. Prescribed fire and managed wildfires are seen as a necessary follow-up treatment [47]; cutting competing vegetation alone increases understory light conditions but does not stimulate the suckering of new ramets as effectively as fire [48]. However, fire extinguishment occurred on two out of three sites under low open wind speeds where canopy fuels were untreated or had lighter cuttings. Interestingly, these two sites had a higher proportion of aspen surface fuels and lowest mid-flame wind speeds. These results mirror the burning experiments of Alexander and Sando [12], who found limited fire spread during prescribed fires in aspen-dominated forests. However, as conifer removal increased in our simulations, fires spread fully across our sites. Heavy cuttings were most effective at permitting fire spread, even though the resulting fires did not result in additional changes to forest structure. Fires can provide several benefits beyond controlling tree density, such as nutrient cycling and surface fuel consumption [5], in addition to stimulating aspen suckering [47]. Coupled with reduced conifer competition, these benefits would promote aspen on our study sites while also impacting their surrounding matrix of conifer-dominated stands. Facilitating restorative fire in these stands may also constrain landscape spread of undesirable high-severity fires while adding to landscape biodiversity [5]. Such effects for example were seen in the Sierra Nevada where four decades of managed wildfires resulted in a landscape mosaic of forest cover types, including the creation of aspen-dominated stands [49].
Over the range of pre-cutting to medium intensity cutting, spread rates increased with additional conifer removal. Increased fire rates of spread are a common outcome following fuel treatments because canopy removal reduces the amount of drag imposed on the wind [50]. However, in the simulations, medium and heavy cuttings altered the mode of fire spread to strictly surface fire with little crown consumption regardless of wind speed. This was due in part to (1) increases in the proportion of aspen fuelbed type in the surface fuelbed, (2) very few large conifer trees with tall crown base heights which were unlikely to torch, and (3) predominance of a less-flammable aspen overstory. This tradeoff of the rate of spread and canopy consumption is not apparent in previous fire modeling efforts within stands containing aspen which independently parameterize the fuels complex and fire weather [11]. The interdependency between fire weather and fuels inherent in fuels treatments demonstrates that reduced surface fuel load and higher crown base heights may compensate for more severe fire weather within the canopy [51].
We do not, however, expect widespread applicability of our findings across aspen restoration treatments in all aspen-conifer forests. Firstly, had cuttings removed trees throughout the canopy strata, rather than from below, mid-flame wind speed increases may have been dampened. Though this implies lower fireline intensities, this silvicultural approach would retain lower residual crown base heights and perhaps maintain some crown fire hazard. Secondly, treatments can also result in considerable additions to surface fuel loads when slash remains on site. This is often the case in remote areas [9] and may offset gains in fire behavior reduction [50,51]. Further experiments employing fuel manipulations in these dynamic ecological systems, coupled with physics-based fire modeling, can address the full suite of silvicultural and fuel management options available to managers [25].

Limitations
Although our study supplied some insight of how the intensity of conifer removal influences stand structure and potential fire behavior and effects, several limitations warrant discussion. First, we utilized a small number of sites (n = 3) in our study; while this did not introduce bias, a broader sample of mixed aspen-conifer stands will increase confidence in extrapolating our results.
Concomitant with the control provided by computational experimentation are decisions on parametrization. As our study was focused on the impact of altered fuels structures on fire behavior, topography was held constant across simulations. We expect that steeper slopes would permit fire spread even under low wind scenarios, leading to increased fire intensity and exaggerating the differences between our cutting scenarios as differences in canopy base heights become a more significant determinant of crown fire activity.
Though our simulations included heterogeneous surface fuels, our fuelbeds were represented using a single set of values for each fuelbed type. This approach is consistent with current operational methods of assessing fire behavior, though it likely resulted in a simplification of the actual spatial and temporal heterogeneity of surface fuels. Additionally, fuel moisture is a particularly important regulator of fire behavior because aspen in the overstory facilitates understory growth whose live fuel moistures can dampen fire behavior [7,52]. As we noted, heavy cuttings had lower rates of spread than medium cuttings because aspen surface fuels were most abundant despite experiencing the highest mid-flame wind speeds. Had we assigned greater fuel moistures to aspen surface fuels, we would have expected to see lower rates of spread after heavy thinning. Lastly, our design captured the expected dynamical turnover of surface fuels associated with conifer removal, though it did not account for any post-cutting slash generated. Leaving such activity fuels on the surface can increase fire behavior and severity, diminishing treatment effectiveness for a period [49]. Research in Sierra Nevada mixed conifer forest suggests activity fuels appreciable decay within seven years of cutting treatments [53]. Thus, our simulations likely underpredicted the short-term potential fire behavior and effects on these sites.
Furthermore, we assumed that post-fire conifer mortality occurred solely as a function of crown consumption. Although this approach inherently simplifies the direct and indirect causes of fire-induced tree mortality [40], previous research has indicated that conifers are likely to die with minimal crown consumption [41,42]. Further development of post-fire tree mortality models which explicitly link species level physiological response to physical properties of heat transfer that can be predicted in physics-based models such as WFDS are needed to allow for a better understanding of linkages between fire behavior and fire effects [54], thus advancing our knowledge of fuel treatment effectiveness [47] and long-term aspen health.

Future Directions
The use of physics-based modeling such as WFDS has emerged as an important tool for advancing our understanding of the complex processes and feedbacks that drive wildfire dynamics and effects [25,55]. In this study, we used a physics-based model to explore the effect of conifer removal intensity on potential fire behavior and effects across three wind speeds. While previous modeling and empirical studies have identified fire weather, surface fuelbed type, and load as critical factors of fire behavior and effects, the ability to explicitly model their interactions is unique to physics-based fire behavior models [55]. These types of models also allow us to investigate fire behavior and effects using virtual experiments. These experiments can serve as a substitute for real-world studies that would be impossible, too costly, time-consuming, or risky. Moreover, the insights from these models can motivate new, targeted experiments and assist in the interpretation of empirical observations [25]. While physics-based models provide an opportunity for advancing our understanding, it is also essential to continuously compare their results with empirical data so that we understand their limitations and uncertainties. Unfortunately, there appears to be a lack of well quantified empirical data on fire behavior in aspen-dominated forests of the Lake Tahoe Basin and, more broadly, the interior western US. To better understand model performance and advance our understanding of fire behavior and effects in mixed aspen-conifer forests, there is a need for additional documentation of fire behavior through further case studies [56] and new experiments such as those discussed by Hoffman et al. [25].  Funding was also provided from the Lake Tahoe Basin Management Unit and through the Southern Nevada Public Lands Management Act as part of science to support the Lake Tahoe West collaborative restoration project.