Study Site and Experimental Design

An invasive shrub removal experiment was initiated in 2009 at Hartley Woods, part of the Arboretum at the Pennsylvania State University. This woodlot is a remnant, mature, mixed-hardwood forest dominated by oak, hickory, and maple with 30–50 cm soil depth to a limestone bedrock substrate (USDA 2020). The experiment includes five sites each with a paired, 20 m diameter treatment (i.e., invasive shrub removal) and control (i.e., invaded) plot (10 plots in total, Kaye and Hone 2016). The invasive shrub removal treatment was first implemented in 2009 and treatments have been maintained since. The removal plots have seen a significant increase in the passive natural regeneration of native shrubs 1–4 m in height (847 individuals ha^–1), but there remains on average 8556 fewer individuals per hectare in this height class in the removal plots compared to plots with invasive shrubs (Maynard-Bean and Kaye 2019). This aligns with expectations for a mature eastern deciduous forest because invasive shrubs create a more closed understory than both recent (e.g., pre-invasion, Woods 1993) and historical conditions (Braun 1916 in Collier et al. 2002). In fact, the native plant abundance in the removal plots exceeds what was expected for native plant recovery based on uninvaded portions of the forest (Maynard-Bean and Kaye 2019). Because of the successful establishment of native understory plants, the removal treatment and control plots are referred to herein as invaded and native for simplicity. The four most abundant native and invasive shrub species and their abundances in unmanaged portions of the forest are listed in Table 1.

Light and Temperature Measurements

Photosynthesis is driven by the number of photons within the spectral range that plants use (400–700 nm), as opposed to the energy contained in each photon. Therefore, PAR is a measure of photons per surface area and time (e.g., µmol m^–2 s^–1). Light was quantified with an LP-80 AccuPAR ceptometer (Decagon Devices, Pullman, Washington) above and below the shrub layer (3 m and 0.1 m from the forest floor, respectively) at the same 20 collection points (two points in each of the five treatment and five control plots) on 27 dates spanning three years (Mar 2016–Jan 2019). Measurements were taken with the ceptometer oriented in the four cardinal directions and averaged for each height at each collection point. Data were collected on sunny days during the 3 hr maximum daily zenith window. Frequently, the maximum zenith for a day was not sunny, resulting in uneven gaps between collection dates (i.e., 1–6 wk). Between measurements in each set of five paired native and invasive shrub plots, PAR measurements were taken at two collection points in the open, outside of the forest canopy, for comparison to the measurements taken in the understory.

Table 1.

The four most abundant native and invasive shrub species with abundances (unmanipulated) at the Arboretum at the Pennsylvania State University's Hartley Woods.

We measured light energy near the forest floor of native and invasive plots in lumens m^–2 (also called lux). It is possible to convert lumens to PAR, but the relationship changes seasonally in an understory and is nonlinear (Long et al. 2012). Furthermore, we wanted to quantify a broader spectrum of light wavelengths that may affect the forest understory beyond PAR. HOBO Pendant Temperature/Light 8K UA-002-08 Data Loggers (Onset Computer Corporation, Bourne, Massachusetts) measure a range of the light spectrum extending into both ultraviolet and infrared light (200–1200 nm). The HOBO loggers were deployed in each of the ten 20 m diameter plots (i.e., five sites each with paired invasive and native plots). Each logger was mounted 0.5 m above the forest floor on a small wooden block attached to a metal stake driven 3 m in a random direction from plot center. To record hourly light (lumens m^–2), one sensor per plot was installed on 3 April 2015 (i.e., 10 loggers). A second sensor per plot was added on 14 October 2017 through 17 January 2019 in the opposite direction from plot center as the first logger (i.e., totaling 20 loggers). When two sensors recorded per plot, the data were averaged by plot prior to further analyses. Two gaps in the data collected within this time frame resulted from the memory filling up more quickly than anticipated omitting most of the summer of 2015 and the spring of 2016. Additionally, from 12 July 2016 through 17 January 2019 one sensor per plot recorded hourly air temperature data mounted on 0.5 m stands with wooden shade structures (to prevent direct sunlight and temperature inflation). The reported resolution of this instrument for temperature is 0.14 °C at 25 °C (0.25 °F at 77 °F), and accuracy is ± 0.53 °C from 0 to 50 °C (± 0.95 °F from 32 to 122 °F).

Phenology Measurements

Plant-level leaf phenology data were collected from January 2016 through February 2019 to understand the influence of the novel, extended phenological window of invasive compared to native shrub layers on the light and temperature data. Phenological data were collected for six individual shrubs of each of the four most common invasive shrub species and four most common native shrub species (Table 1) using standardized protocols from the USA National Phenology Network outlined below (Denny et al. 2014). The shrubs were located in unmanaged areas among the five sites within the forest stand containing the invasive shrub removal experiment. The presence of leaves is defined as at least one base of a leaf, specifically the point of attachment with the petiole, emerged from a bud on a given plant. Dried or dead leaves are not considered as part of a “yes” for leaves present. The percentage of the potential canopy space is estimated categorically: less than 5%, 5–24%, 25–49%, 50–74%, 75–94%, 95% or more. Here, we consider the phenological window to begin and end when the canopy increases and decreases to 75–94% of full canopy, respectively. Each sampled shrub, 48 individuals, was visited at least twice per week during the beginning and end of the growing season and at least once per two weeks in the middle of the growing season.

Data Analyses

PAR data collected 3 m above the forest floor were used to model seasonal PAR available to the shrubs in the understory, generalizing point measurements to a continuous seasonal pattern. As expected, these data follow a sinusoidal curve across the collection years that aligns by day of year. Therefore, a fourth-order polynomial regression of the day of year (DOY) of data collection with year (year) and collection site (site) as random effects was fit as in Equation 1:

Here, PAR (µmol m^–2 s^–1) is light above either the invaded or native shrubs. All data analyses and figures herein were performed in the programming environment R 3.4.4 (R Core Team 2017). The above mixed-model was fit using the lmer function (lme4 package, Bates et al. 2015) where β0, β1, β2, β3, and β4 are the leading coefficients of the polynomial function, ε is residual error, and σ, σ_s, and σ_y represent the associated variance terms. The assumptions surrounding normality and homoscedastic variance were tested graphically using base plotting functions in R. The model estimates were tested using Satterthwaite's method for approximating degrees of freedom for mixed models (lmerTest package, Kuznetsova et al. 2017). To quantify the variance in the data explained by the model including both fixed and random effects, the conditional coefficient of determination (hereafter, R²) value was determined using the r.squaredGLMM function (MuMIn package, Barton 2018). To best represent the estimated fit line and to calculate the associated 95% bootstrapped confidence intervals (CIs), the predict function was used within the bootMer function to produce 1500 bootstrapped estimates of PAR available to the shrubs for each day of the year (boot package, Canty and Ripley 2017). To test whether the randomly assigned treatment and control differ in light infiltrating the overstory canopy, the same model was fit for the data split by native and invasive plots. There was no difference in the relationships, verifying no overstory canopy bias by treatment (results not shown).

Outside of the forest canopy there is a seasonal pattern to the light received from the sun at a given latitude. Seasonal light above the overstory tree canopy (i.e., in the open) was measured to contextualize light availability above the shrubs and near the forest floor. This was modeled using a second order polynomial with year as a random effect as in Equation 2:

Here, PAR (µmol m^–2 s^–1) is the light in the open. The above mixed model was fit using the lmer function (lme4 package, Bates et al. 2015) where β0, β1, and β2 are the leading coefficients of the polynomial function, ε is residual error, and σ and σ_y represent the associated variance terms. The assumptions and model effects were tested, R² calculated, and relationships and CIs estimated as described for Equation 1 models, above.

The photosynthetically active radiation (PAR, µmol m^–2 s^–1) measured at 3 m and 0.1 m was used to calculate the amount of PAR filtered by the shrub layer at a given sampling point (i.e., above minus below). The difference in PAR filtered between each of the five paired plots (i.e., invasive minus native) provides the light filtered out by invasive shrubs as compared to or relative to the native shrub layer. To provide a context for this metric, we also calculated percent reduction in PAR filtered by dividing the difference by the light below native shrubs. As no relationship adequately fit these relatively noisy time series data in terms of significant estimates that did not violate test assumptions related to normality, the data were bootstrapped using the boot package (Canty and Ripley 2017). Bootstrapping with replacement (1000 replicates) across sites for each of the 27 sampling dates (n = 5) was used to estimate means as well as the “percentile” 95% CIs appropriate for data that diverge from normality and data of small sample size (Davison and Hinkley 1997; Carpenter and Bithell 2000). These mean values of PAR filtered and percent reduction in PAR for each date were bootstrapped by the four meteorological seasons: spring (March through the end of May), summer (June through August), fall (September through November), and winter (December through January). The PAR filtered and percent reduction for the dates specifically falling during the ELP of invasive shrubs compared to native shrubs was also calculated (i.e., spring ELP, fall ELP). The mean effect of invasive shrubs on light filtered by the shrub layer can be tested for significance (α = 0.05) by 95% confidence intervals (CIs) that do not overlap with zero.

The light intensity data (lumens m^–2) was recorded hourly at 0.5 m above the forest floor. The difference in daily average intensity between the paired invasive and native plots was calculated to provide the effect of invasive shrubs on light infiltration (i.e., invasive minus native). To provide a context for this metric, the percent reduction in light was calculated by dividing the difference in infiltration by the light intensity below native shrubs. These data contained too much variability to fit a reliable relationship with significant estimates that did not violate test assumptions, similar to differences and percent reduction in PAR. The difference in light energy available below invasive relative to native shrubs and the percent reduction in light energy for the five paired plots were bootstrapped with replacement (1000 replicates) by calendar week (n = 35). The weekly values were used to calculate seasonal means and confidence intervals following the methods described for the effect of invasive shrubs on PAR filtered by the shrub layer, above.

Figure 2.

Light filtered out by invasive relative to native shrub layers with phenological windows. The vertical background bands represent the seasons as indicated in Figure 1. The black points are the bootstrapped mean difference in light intercepted by the shrub layer for each sampling date with 95% confidence intervals (n = 5). The confidence intervals are connected by transparent gray bands for visualization purposes. Black confidence intervals that overlap the dotted line (y = 0) indicate that invasive and native shrubs do not differ in the amount of light filtered for a given sampling date. If the confidence interval is entirely above the dotted line, invasive shrubs filter more PAR than native shrubs. The gray and white rectangles represent the phenological windows for invasive and native shrubs, respectively, with the black bars indicating 95% CIs around dates of emergence and senescence. The model estimates and coefficients for phenology data are available in   Supplementary Content (186_naar-41-03-02_s01.pdf) at BioOne online.

Temperature data (°C) was also recorded hourly, and daily minimum and maximum values were summarized. To understand the influence of invasive shrubs on the daily minimum and maximum temperatures, the difference between the paired invasive shrub plots and native plots was calculated (i.e., invasive shrubs minus native shrubs). The daily difference in minimum and maximum temperatures were each bootstrapped with replacement (1000 replicates) by calendar week following the methods described above for the effect of invasive shrubs on PAR filtered by the shrub layer.

Table 2.

The seasonal influence of invasive shrubs on light relative to native shrubs.

Differences in leaf phenology by species and origin (native vs. invasive) were estimated using a nested ANOVA framework for day of year (DOY) of leaf emergence and DOY of leaf off for each year. The models were fit as in Equation 3:

Here, one model was fit for each DOY leaf emergence and leaf off and for each year with µ as the overall mean (or common value of DOY), α as the effect of origin, β as the effect of species nested within origin, ε as residual error, and σ represents the associated variance term. The model was fit using the lm function, which provides tests of the model estimates and a model R² value. The assumptions were tested as described for the Equation 1 model above. Because of the nested effects, effects were tested and group means determined using estimated marginal means, also known as least-squares means (emmeans package, Lenth 2018).

Light and Phenology

Light and phenology in the understory are constrained by the phenology of the deciduous forest overstory (Martin et al. 2009). In the open, away from the forest canopy, the annual maximum PAR peaked between June and July, while below the tree canopy and above the shrub layer the peak occurred between March and April at 45% of the peak in the open (Figure 1, model estimates and coefficients available in  Supplementary Content (186_naar-41-03-02_s01.pdf) at BioOne online). Invasive shrubs either filter more PAR than natives at each sampling date (40.7% of the 27 sampling dates), or there is no difference in the amount of PAR filtered by the shrub layers (59.3% of the 27 sampling dates, Figure 2). On average, invasive shrubs filter 58.36 [38.46, 78.74] µmol m^–2 s^–1 more PAR than native species corresponding to a 39.4% reduction in PAR relative to native shrubs. This supports the first part of the hypothesis (H₁) that less light will infiltrate the invasive shrub layer, but the ELP of invasive shrubs compared to native shrubs in the spring appears to filter greater PAR than in the fall (Figure 2, 95% CIs during ELP are above zero in all three springs, but only one of two falls). However, the average PAR filtered during the ELP windows does not significantly differ from other seasons (i.e., 95% CIs of ELP overlap with CIs of other dates, Table 2). Relatively wide confidence intervals reflect high heterogeneity in understory light availability along with relatively small sample sizes. Still, we find that invasive shrubs filter more PAR than natives across seasons, indicating that native shrubs do not produce as much shade as invasives even when they both have leaves in the summer (Figure 2). Furthermore, differences that occur in the winter months indicate that even leafless branches of invasive shrubs filter more PAR than the native community. Together this verifies observations of greater foliage density or leaf area index for invasive compared to native shrubs in the understory of deciduous forests (e.g., Woods 1993, Maynard-Bean and Kaye 2019). The range of PAR available below invasive shrubs was great (3–988 µmol m^–2 s^–1), and rarely approached the light compensation point for understory herbs (e.g., 5–16 µmol m^–2 s^–1, Hull 2002) and woody seedlings (e.g., 5.3–9 µmol m^–2 s^–1, Craine and Reich 2005) common to eastern North American deciduous forests. Even so, the negative effect of invasive shrubs on PAR that we found here has the potential to dramatically influence the carbon budget of shade-adapted understory species (Chazdon and Pearcy 1991).

Figure 3.

Light energy available below invasive relative to native shrubs. The vertical background bands represent the seasons as indicated in Figure 1. The black points and corresponding error bars are the bootstrapped weekly mean difference of the daily average light intensity below invasive and native shrubs with associated 95% CIs (n = 35). The intervals are connected by the dark gray bands for visualization purposes. Intervals that overlap the dotted line (y = 0) indicate that the invasive and native shrubs do not differ. If the confidence interval is entirely below the dotted line, there is significantly less light available below invasive shrubs as compared to native shrubs. Table 2 shows the bootstrapped means and 95% CIs for seasonal differences.

In addition to PAR, we also measured light energy (lumens m^–2) spanning a larger range of the electromagnetic spectrum. The light energy of invaded plots is significantly reduced in 60.6% of weeks sampled from 2015 to 2019 (i.e., 106 of 175 weeks, Figure 3). Seasonally, invasive shrubs allow significantly less light infiltration compared to native shrubs. The largest difference in light infiltration occurs in the spring (–1981 [–2604, –1380] lumens m^–2), with the magnitude of difference decreasing each season through summer, fall, and winter (Table 2). This seasonal trend corresponds to light availability in the understory (Figure 1), and when taken relative to understory light availability as the percent reduction in light there are not differences by season (Table 2). So, despite a similar percent reduction in light energy, mean light energy is much lower during the spring ELP of invasive shrubs (–2809 [–3500, –2109]) than for fall ELP (–722 [–841, –570]). Spanning nearly four years, light infiltration was generally reduced by the presence of invasive shrubs. Similar to PAR, this lends further support to observations of greater foliage density for invasive compared to native shrubs in the understory of deciduous forests. Decreased understory light energy negatively influences the movement and behavior of invertebrates (McKinney and Goodell 2010), birds (Bennett and Cuthill 1994), and amphibians (Enright et al. 2015) and has the potential to incur indirect impacts to species and ecosystem processes through reductions in the transfer of light energy to heat at the forest floor.

Temperature

Due to the novel shade produced by the invasive shrub canopy, we hypothesized (H₂) lower/colder daily maximum temperatures. Invasive shrubs decreased the maximum temperature in the summer and fall of 2016, and the spring and summer of 2017, but there was no effect on air temperature for 2018 (Figure 4b). This may be related to abnormally wet weather in 2018. Compared to the average of 2016 and 2017 (which aligned with 1981–2010 climate normals), the mean monthly maximum air temperature was 1.2 °C lower and precipitation was 70% greater in 2018 (NOAA 2020). The high specific heat of water means that it requires a lot of energy to change temperature compared to the rest of the forest floor. A moist understory in 2018 likely influenced the transfer of light energy to heat. Nevertheless, this pattern of invasive shrubs reducing maximum air temperature during the growing season of two years strengthens existing knowledge of this effect observed for one invasive shrub species (L. maackii, Watling et al. 2011, Chen and Matter 2017). Because of the potential insulative effects of a novel shrub canopy, we hypothesized (H₃) higher/warmer daily minimum temperatures. Minimum air temperatures were not clearly influenced by the novel shade produced in the understory by invasive shrubs. The coarse resolution of temperature measurements (± 0.5 °C) may not detect the fine resolution differences that invasive shrubs could cause. Additionally, a small difference in air temperature between invasive and native plots could be masked by air movement in the forest understory. Even with the large sampling plots used here (20 m in diameter) as well as the placement of sensors within 3 m of plot center, the air temperature of the regenerated native community is likely still influenced by the surrounding matrix of invasive shrubs. Placing multiple sensors in the upper soil layers may ameliorate this effect because soil temperature changes more slowly than air temperature.

Figure 4.

The influence of invasive shrubs on daily minimum and maximum air temperature compared to native shrubs. The background gray bands represent the seasons as indicated in Figures 1–3. The darkened areas indicate the bootstrapped weekly 95% confidence regions of the (a) median understory forest temperature across date, and the difference between the (b) maximum and (c) minimum air temperature around invasive compared to native shrubs. The black points and corresponding error bars are the bootstrapped weekly mean difference in the daily (b) maximum or (c) minimum air temperature in the understory with associated 95% CIs (n = 35). Black CIs that overlap the dotted line (y = 0) indicate that the invasive and native shrubs do not differ for a given weekly value. If the confidence interval is entirely above the dotted line, invasive shrubs increase that temperature value compared to natives. If below, invasive shrubs decrease temperature compared to natives. However, results less than about 0.5 °C in difference should be interpreted with caution (instrument accuracy ± 0.53 °C from 0 to 50 °C).

Significance

In the deciduous forest understory, there is less light energy (lumens m^–2) available below invasive shrubs than below native shrubs with mean differences largest in the spring and differences decreasing through summer, fall, and winter. Likely related to light energy reductions, maximum air temperature was reduced during the growing season of two years. Invasive shrubs filter more available PAR than natives, and seasonal differences were not detected. Light is an important resource in the understory of deciduous forests, and modest changes have the potential for dramatic consequences to carbon budgets (Chazdon and Pearcy 1991). Also, the light environment alters nutrient allocation for shade-tolerant understory species, with more nitrogen allocated to leaves under lower light conditions (Niinemets 1997). This has potential implications for increased resource competition between native understory plants and invasive shrubs, and for litter quality and decomposition rates. The darker, cooler conditions with shrub invasion have the potential to exacerbate the mesophication of eastern deciduous U.S. forests—a phenomenon that has resulted from a positive feedback loop of fire suppression allowing mesophytic, shade-tolerant native species to shift the understory toward more cool, damp, and shaded conditions resistant to burning (Nowacki and Abrams 2008). Additionally, quantifying the changes to seasonal light could allow adjustments to models predicting native species populations or ecosystem processes. For example, the influence of light on germination rates are known for many plant species (Baskin and Baskin 2001) and could be used to modify demographic models to estimate changes to population growth rates. Furthermore, the changes to light quantified here can be used to experimentally manipulate understory light availability or growth chamber conditions to disentangle the influence of the novel shade of invasive shrubs on, as examples, seed production, plant growth, and invertebrate transit, development, or emergence. The presence of invasive shrubs reduces average PAR availability by 39.4% [25.4, 54.5] compared to an abundant native understory community with seasonal influences on light energy and temperature near the forest floor, which can negatively impact native species and ecosystem processes.