key: cord-0729712-zvk2438q
authors: Young, Alyssa L.; Bloodworth, Kathryn J.; Frost, Morgan D. T.; Green, Curtis E.; Koerner, Sally E.
title: Heatwave implications for the future of longleaf pine savanna understory restoration
date: 2021-11-26
journal: Plant Ecol
DOI: 10.1007/s11258-021-01212-7
sha: d15936f39b8794a9bb884c1a331a9087127312df
doc_id: 729712
cord_uid: zvk2438q

The longleaf pine (LLP) savanna ecosystem once covered ~ 92 million acres of the Southeast USA, but due to anthropogenic activities such as logging and fire suppression, only 3% of its once widespread historic range remains. While many restoration efforts are underway to conserve this biodiverse ecosystem, restoration must be done in the context of climate change. In the last few decades, heatwaves have increased in frequency and intensity across the Southeastern USA with further increases predicted. To expand our understanding of LLP savanna restoration in light of these changes, we ran a series of three simulated heatwave greenhouse experiments through a Course-based Undergraduate Research Experience (CURE) incorporating ~ 150 undergraduate researchers per experiment. We measured plant growth metrics for four understory grasses commonly used in LLP savanna restoration efforts. We found that while most grass plug individuals survived heatwave conditions, aboveground production was reduced due to heatwaves. This productivity decrease could result in less biomass available for the essential vegetation fire feedback loop, where fire increases grass biomass, and in turn, more grass provides more fuel for fire. These results imply that land managers can proactively compensate for biomass loss due to heatwaves by planting more grass plugs during initial restoration. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s11258-021-01212-7.

Restoration of ecosystems is a key priority in biodiversity conservation globally. However, restoration success is threatened by climate change (Lavendel 2003) . Worldwide, anthropogenic climate change has resulted in global warming and increased variability in the hydrological cycle (Hayhoe et al. 2018) . As a result, many regions around the world are experiencing higher occurrences of extreme weather events, such as heatwaves (Overpeck 2013) ; (Mazdiyasni and AghaKouchak 2015) ; (Ades et al. 2020) , here defined as increased temperature accompanied by decreased precipitation and humidity. Therefore, to properly restore the biodiversity in degraded ecosystems, climate change must be factored into current and future restoration plans (Harris et al. 2006 ); (Wilsey 2020). When climate change is accounted for in restoration planning, practitioners will be better equipped to work within an adaptive management framework to more successfully restore ecosystems and, therefore, conserve biodiversity (Buckley et al. 2011) .

Within the USA, the longleaf pine (LLP) savanna ecosystem in the Southeast is of particular importance when considering the intersection of restoration and climate change (Noss et al. 2015) ; (Kirkman and Jack 2017). Currently, there are large efforts underway to restore the LLP savanna ecosystem (Oswalt et al. 2012) ; (Holland et al. 2019) . Prior to the European settlement of North America, the fire-dependent LLP savanna ecosystem covered an estimated 92 million acres of the Southeast USA. However, anthropogenic disturbances, including fire suppression, led to the degradation and loss of the ecosystem and, therefore, much of its biodiversity (Frost et al. 2006) . Today, the LLP savanna ecosystem covers only 3% of its historic range (Oswalt et al. 2012) . While the overstory of LLP savannas are dominated by a single tree species, the longleaf pine tree (Pinus palustris), the understories of LLP savannas are rich in graminoid and forb species, with some sites having [ 40 vascular plant species in a 1m 2 area (Walker and Peet 1984) ; (Brudvig et al. 2013) . Therefore, restoration that focuses on the understory plant community is needed (Veldman et al. 2015) . However, the Southeast USA is also experiencing increased heatwaves that are expected to worsen as the climate continues to change (Hopkinson et al. 2013) ; (Hayhoe et al. 2018) ; (Hoegh-Guldberg et al. 2018) . The practical tools for LLP understory restoration are underdeveloped, particularly in the face of climate change.

While the understory of LLP savannas is important from a biodiversity perspective, it is also critical to the sustainability of restoration projects due to its role in the vegetation-fire feedback loop. LLP savanna is a fire-dependent ecosystem, requiring burning to reduce litter on the understory floor and eliminate the competitive advantage of woody species (Brockway et al. 2006) ; (Walker et al. 2006) ; (Gagnon et al. 2015) . Frequent burning in LLP savannas increases grass dominance and biodiversity, while maintaining an open canopy (Barnett 1999) ; (Brockway et al. 2006) ; (Veldman et al. 2014) . The productivity of the grass layer interacts with fire in a positive feedback loop-a productive understory leads to a high fuel load, the high fuel load leads to frequent fire, and frequent fire leads to a productive understory (Brockway et al. 2006) ; (Barlow et al. 2012) ; (Fill et al. 2016 ). This vegetation-fire feedback loop is a critical target for restoration efforts of this savanna ecosystem; without a productive grass layer to fuel fires, the system cannot persist long term without intensive, costly, and continuous management in place.

Unfortunately, the understory grasses and forbs are also sensitive to climate change impacts, in part due to more shallow roots than their longleaf pine tree counterparts (Clewell 1989) ; (Beckage et al. 2006) ; (Brockway et al. 2006) . Numerous studies have shown grassland communities to be resilient to heatwaves, with decreased production initially, followed by a quick recovery (e.g., (Weaver 1954) ; (Kreyling et al. 2008) ; (Hoover et al. 2014) ; (Isbell et al. 2015) . However, in a restoration context, newly planted grass plugs may be less resilient (Stromberg and Kephart 1996) ; (Huddleston and Young 2004) . Additionally, heatwave interactions with disturbance could lead to a decrease in grass dominance (Ratajczak et al. 2019) , impacting fuel load. Thus, climate change, particularly in fire prone ecosystems, has the potential to make the restoration of the vegetation-fire feedback loop unsuccessful and must be studied in more depth. To understand the impact that climate change in fire prone ecosystems has on the vegetation-fire feedback loop, we must first assess how different understory grass species, important in the vegetation-fire feedback loop in the LLP savanna, react to heatwaves in their first year after planting.

Here, we conducted a series of three separate greenhouse experiments (Exps) to understand how heatwaves influence the growth and development of native grass plugs commonly used in LLP savanna restoration. One heatwave event was simulated during each experiment by increasing heat and decreasing humidity and precipitation for half of the grass plugs. Heatwave severity and grass species differed slightly between each experiment, giving each an independent experimental approach. However, the unifying goal is to understand how the occurrence of a heatwave influences the restoration of understory grass species and, therefore, the vegetation-fire feedback loop. Utilizing three independent experiments allowed us to explore the effects of heatwaves within three separate contexts, increasing the robustness of the shared findings. Across all three experiments, we measured weekly plant growth metrics (maximum leaf length, maximum plant height, and stem number) and at the end of each experiment, we measured leaf area. Using this information, we determined the heatwavesensitive morphological traits that lead to biomass changes; biomass was measured at the end of each experiment, with aboveground biomass measurements for all three experiments and belowground biomass measurements for Exp 2 only. While we hypothesized that growth would be negatively impacted by the occurrence of a heatwave, our study aimed to explore the relative sensitivities of morphological traits that lead to reduced growth within and across species.

We conducted a series of three consecutive greenhouse experiments (Table 1 ) exploring the effects of heatwaves on native grass plug growth. We performed all three experiments in the Biology Department's Teaching Greenhouse at the University of North Carolina at Greensboro (UNCG), USA. All three experiments were part of a Course-based Undergraduate Research Experience (CURE) in the Ecology and Evolution Lab required for all UNCG Biology undergraduates. We conducted one experiment per semester, with approximately 125 undergraduate researchers per experiment participating in experimental set up, data collection, and preliminary data analyses.

We obtained grass plugs of four native species common in LLP understories from the North Carolina (NC) Forest Service Claridge Nursery, a nursery that sells grass plugs specifically for LLP savanna restoration across the state of NC. Across our three experiments, we used Andropogon gerardii, Aristida stricta, Schizachyrium scoparium, and Sorghastrum nutans. All four are highly productive, C4, perennial grasses which contribute substantially to the vegetation-fire feedback necessary for healthy LLP savanna. We trimmed the bottom * 2.5 cm of all grass plugs and gently loosened the roots before potting using SunGro 3B Potting Mix in plastic pots (12.7 cm 9 8.3 cm 9 8.3 cm). Undergraduate researchers planted all of the grass plugs into separate pots (i.e., one grass plug per pot), and grasses were then placed in trays located on benches in the greenhouse, where temperature, humidity, and light availability were monitored throughout the experiment (Fig. 1) . We then assigned each pot to one of two treatments: control or heatwave. We watered plugs to maintain soil moisture above the wilting point, except for the heatwave grass plugs during the simulated heatwave. All plugs were always in identical conditions, except during the heatwave. To account for minor microclimatic differences in the greenhouse, we alternated control and heatwave grasses across the bench when the heatwave was not being imposed. To simulate the heatwave, we transferred half of the grass plugs to a different bay of the greenhouse where humidity was substantially lower (45% in Exp 1, 69% in Exp 2, and 51% in Exp 3; Fig. 1 ), temperature was higher (14% in Exp1, 20% in Exp 2, and 39% in Exp 3; Fig. 1 ), and the plugs received no water resulting in reduced soil moisture (42% in Exp 1, 97% in Exp 2, and 67% in Exp 3; Fig. 1 ). The duration of the heatwave treatment for each experiment differed, with heatwave plugs experiencing a 15-, 11-, and 9-day heatwave for Exps 1, 2, and 3, respectively (Table 1) . At the end of the heatwave, we moved heatwave plugs back to the bench with the control plugs, returning them to equivalent conditions. We determined the length of the heatwaves for each experiment based on several factors. Single heatwave events in the Southeastern United States are predicted to last from 5 to 20 days under predicted global climate temperature changes (Perkins-Kirkpatrick and Gibson 2017), so we chose heatwave lengths in the greenhouse to reflect this variability, with a target of 10 days. Additionally, altering the temperature in the greenhouse partially depends on outside air temperature. Thus, the heatwave in Exp 1 was longer than 10 days due to low outside temperatures and Exp 3 was shorter than 10 days due to warm outside temperatures. Finally, as these experiments were conducted as part of an undergraduate course, we were confined to each semester's schedule and adjusted the heatwave accordingly to maximize student's measurements while allowing time to properly analyze and report their results. Undergraduate researchers measured maximum leaf length, plant height, and stem number weekly throughout the experiment over 68, 68, and 66 days during Exps 1, 2, and 3, respectively (Table 1) , with primary production and leaf area measured upon completion of the experiment, which was 25, 11, and 1 day after the end of the heatwave treatment for Exps 1, 2, and 3, respectively (Table 1) . For maximum leaf length, undergraduate researchers measured the longest green leaf by placing the ruler gently on the soil surface at a 90°angle to the soil and extending the longest green leaf up along the ruler to its full length. Researchers measured plant height as the height of the tallest green part of the plant without manipulating the grass leaves. Undergraduate researchers counted stem number as the number of green stems sprouting from Mean and standard error of weekly a-c temperature, df humidity, and g-i soil moisture for each of the three independent experiments for control (purple solid line) and heatwave (red dashed line) plants. Blue vertical lines denote the start and end of the heatwave. Significant differences between treatments using a Wilcoxon rank sum test for each experiment, measurement, and week are denoted with * and represent p \ 0.05, p \ 0.01, and p \ 0.001 for one, two, and three astricts, respectively, from paired t-tests. Importantly, p-values were adjusted for multiple comparisons using the Benjamini-Hochberg method before determining significance. Bars represent means ± standard error the base of the plant. We measured aboveground primary productivity by harvesting all aboveground biomass at the end of the experiment and separating it into alive and dead components. Then, we calculated leaf area (cm 2 ) using a leaf area meter (Li-Cor) directly after harvest of aboveground biomass by randomly selecting one fully emerged, live, leaf blade from the alive biomass of each sample and running it through the meter. Then, we dried both the alive and dead biomass for each sample separately in a forced air oven at 60°C for 48 h before weighing. We report aboveground production as the sum of all living and dead aboveground biomass, an indicator of fuel load, created per plug. In Exp 2 we collected belowground productivity for each pot by first sieving the soil and then gently washing the roots. We placed roots into a forced air oven at 60°C for 48 h before weighing. While planned for Exp 3, COVID-19 prevented the ability to gather belowground productivity data, as the experiment ended earlier than planned when UNCG announced research shutdown in March 2020. Undergraduate researchers recorded humidity, temperature, light availability, and soil moisture multiple times per week near mid-day. Researchers monitored humidity and temperature using digital sensors placed among the grasses on each half of the bench. Researchers measured light availability (lux) over each tray of pots using small, handheld light meters (Fisher Scientific). Undergraduate researchers collected soil moisture using a probe (6 inch, Campbell Scientific) which was inserted into ''reference'' pots which contained a grass plug, but these plugs were not monitored for growth due to probe disturbance. Researchers measured all four abiotic variables each time they collected data throughout the semester.

Since undergraduate researchers were mostly new to plant research, and to ensure accuracy and effective experimental manipulations, we put several quality control measures into place. First, a full-time graduate teaching assistant was in the greenhouse at all times with the undergraduate researchers to answer questions and monitor data collection techniques. Second, undergraduate researchers worked in pairs. And finally, two separate groups measured each plant. This allowed us to check the data post collection; we determined the percent difference between the two pseudo-replicated measurements. If measurements were similar, we took the mean. If measurements were very different, further exploration occurred (i.e., original data sheet checked for data entry errors, compared previous and following weeks measures) to determine which record was correct. If no obvious solution could be determined, we dropped measurements for that plant in that week. In addition, at times, data were not submitted at the end of the semester, and was essentially lost, thus replication throughout the experiment varied. This was particularly true for Exp 2. Here we present all data that we were able to obtain and that passed our quality control checks.

We conducted all data manipulation and statistical analyses in R (R Core Team 2020). All variables (both abiotic and biotic) were measured weekly throughout the experiments. However, only the measurements from the last week of the heatwave treatment were used in analyses for biotic variables that were measured weekly. This single week was chosen for the biotic variables for the sake of clarity in assessing multiple experiments across time and so that the same timepoint in each experiment could be assessed, as Exp 3 ended two weeks early due to COVID shutdowns. Importantly, the same timepoint was used for the paired control and treatment plots; therefore, they are directly comparable. Conversely, for the abiotic variables, we assessed data from all weekly timepoints, as this weekly information is important to show the effectiveness of the heatwave treatments and to show that when not in the heatwave, the two treatments were exposed to identical conditions. As our data were not normal and sample sizes were low, we determined differences between control and heatwave grass plugs and abiotic conditions by using a Wilcoxon rank sum test with separate tests run for each experiment, variable, and species. P values were adjusted for multiple comparisons using the Benjamini-Hochberg method (Benjamini and Hochberg 1995) . In Exp 1, p-values were adjusted for the five biotic variables tested within a species, and for the abiotic variables, p-values were adjusted for the ten weekly measurements tested. As a note, there was not enough data to run a Wilcoxon rank sum test on week 8 of the humidity and air temperature data and week 10 of the air temperature data. Therefore adjustments were made based off the total number of weekly measurements where a Wilcoxon rank sum test was performed. In Exp 2, p-values were adjusted for the six biotic variables tested, and for the abiotic variables, pvalues were adjusted for the 10 weekly measurements tested. In Exp 3, p-values were adjusted for the six biotic variables tested, and for the abiotic variables, pvalues were adjusted for nine weekly measurements tested (no data were collected on week 8 of the experiment).

The simulated heatwaves increased temperature and reduced humidity and soil moisture during the treatment, with a lagged recovery following the end of the heatwave (Fig. 1) . Specifically, in Exp 1, we simulated the heatwave from the start of week 5 through week 7 of the experiment. In week 5, the average air temperature was significantly higher in the heatwave (Fig. 1a) . In weeks 5 and 6 the average humidity was significantly lower in the heatwave (Fig. 1d) , and in weeks 5, 6, 7, and 9, soil moisture was significantly lower in the heatwave (Fig. 1g) . In Exp 2, we simulated the heatwave from the end of week 7 through week 9. In week 8, the average air temperature was significantly higher in the heatwave (Fig. 1b) , and in weeks 8 and 9, the humidity and soil moisture was significantly lower in the heatwave (Fig. 1e, h) . In Exp 3, we simulated the heatwave from the end of week 8 through week 10. In week 9 and 10, the average air temperature was significantly higher, while humidity and soil moisture were significantly lower in the heatwave (Fig. 1c, f, i) . Across all three experiments, only two of 29 weeks (Exp 3-weeks 9 and 10) showed significant differences in light availability (lux) between heatwave and control plants (data not shown).

Overall, the fifteen-day heatwave of Exp 1 (Table 1 ; Fig. 1 ) infrequently caused plant death (Table S2) , but tended to decrease measures of plant growth; however, significant differences between treatment and control growth metrics were not common (Fig. 2 & S5 ; Table S3 ). In fact, for three of the four species (A. gerardii, S. scoparium, and S. nutans), no variable measured was significantly different between the two treatments ( Fig. 2 & S5 ; Table S3 ), even though aboveground biomass in Week 9 was 20%, 35%, and 21% lower (not significant) for A. gerardii, S. scoparium, and S. nutans, respectively ( Fig. 2 ; Table S3 ). A. stricta had the lowest aboveground biomass change of only 12% in Week 9 due to the heatwave, which was not significant (Fig. 2 ;  Table S3 ); however, maximum plant height was significantly decreased in Week 7 by the heatwave by 8% ( Figure S5 ; Table S3 ). We also conducted analyses for the first week post heatwave (Week 8) and saw the same significance trends as mentioned previously (data not shown).

Similarly, the eleven-day heatwave of Exp 2 ( Table 1 ; Fig. 1 ) did not cause plant death (Table S2 ) but significantly decreased measures of plant growth for S. nutans (Fig. 3 & S5 ; Table S4 ). In Week 9, at the end of the heatwave, we saw significantly lower maximum plant height (22%) and stem number (24%). In the last week of the experiment (Week 11), leaf area ( Figure S5 ; Table S4 ) was 58% lower in heatwave grasses compared to control, and production was significantly lower for the heatwave plants compared to the control plants ( Fig. 3 ; Table S4 ). Aboveground production was 16% lower in the heatwave treatment compared to the control ( Fig. 3a; Table S4 ), while belowground production was 25% lower (Fig. 3b; Table S4 ).

Finally, the nine-day heatwave of Exp 3 (Table 1 ; Fig. 1 ) did not cause plant death (Table S2 ) but significantly decreased measures of plant growth for A. stricta (Fig. 4 & S5 ; Table S5 ). We took all Fig. 2 Aboveground production (sum of all living and dead aboveground biomass) from Experiment 1 was not significantly reduced under heatwave conditions for any of the four grasses (Andropogon gerardii, Aristida stricta, Schizachyrium scoparium, Sorghastrum nutans). Aboveground production was collected from Control (C; purple) and Heatwave (H; red)exposed grass plugs at the end of the experiment, 9 weeks after initial planting and 3 weeks after the heatwave ended. Significant differences between treatments were determined by using a Wilcoxon rank sum test for each species. Significance at p \ 0.05, after correcting for multiple comparisons using the Benjamini-Hochberg method, is denoted with an *. Bars represent means ± standard error measurements the week after the heatwave ended (Week 10), as the UNCG shutdown caused by COVID-19 required us to end our experiment 2 weeks earlier than planned. Aboveground production was 13% lower in the heatwave plants compared to the control plants ( Fig. 4; Table S5 ). Similarly, we saw significant reductions in maximum plant height (7%) and leaf area (25%), while maximum leaf length and stem number were not significantly different (Figure S5 ; Table S5 ).

This series of greenhouse experiments shows that an increase in heatwave occurrence, predicted for much of the LLP range (Hayhoe et al. 2018) ; (Hoegh-Guldberg et al. 2018) , will likely have consequences for the restoration of biomass production, with implications for the vegetation-fire feedback loop in LLP savanna restoration efforts. While nearly all individuals survived the heatwave, generally, the heatwave plugs had lower aboveground biomass compared to the control plugs at the end of each experiment. The decrease in aboveground biomass will correspond to less fuel available to carry fire, which is essential to maintaining plant biodiversity. Without a productive understory grass layer, restoration efforts aimed at high understory diversity could be more likely to fail. However, this can be overcome by considering the effects that heatwaves will have on grass plug growth when forming restoration plans.

Overall, we show that the occurrence of a single heatwave decreased growth of commonly found native understory grass species of LLP savannas in some experimental contexts. Across all three experiments and all grass species studied, we found that the occurrence of a heatwave negatively impacted measures of plant growth, though this was most obvious when looking at aboveground biomass. S. scoparium appears to be the most sensitive to heatwaves as seen in Exp 1, with a 35% decrease in aboveground biomass due to the heatwave. While no other species showed significant decreases in biomass in Exp 1, this may have been due to the lower magnitude of response. While variability of response among the grass species was equivalent, the lower magnitude of response of Fig. 3 a Aboveground and b belowground production (sum of all living and dead biomass) from Experiment 2 were both significantly reduced in Sorgastrum nutans, Indian grass, under heatwave conditions. Production was collected from Control (C; purple) and Heatwave (H; red)-exposed grass plugs at the end of the experiment, 11 weeks after initial planting and 3 weeks after the heatwave ended. Significant differences between treatments were determined by using a Wilcoxon rank sum test. Significance at p \ 0.05, after correcting for multiple comparisons using the Benjamini-Hochberg method, is denoted with an *. Bars represent means ± standard error Fig. 4 Aboveground production (sum of all living and dead aboveground biomass) from Experiment 3 was significantly reduced in Aristida stricta, Wiregrass, under heatwave conditions. Production was collected from Control (C; purple) and Heatwave (H; red)-exposed grass at the end of the experiment, 10 weeks after initial planting and the week after the heatwave ended. Significant differences between treatments were determined by using a Wilcoxon rank sum test. Significance at p \ 0.05, after correcting for multiple comparisons using the Benjamini-Hochberg method, is denoted with an *. Bars represent means ± standard error some grasses was not significantly detectable with an n = 24. In Exp 2 and 3 where replication was 4 times higher (n = 96 per treatment) than Exp 1, S. nutans and A. stricta both showed similar magnitudes of biomass decrease as in Exp 1, but in Exp 2 and 3, the decrease was significant. A. stricta appears to be more resilient than the other three species, as we found only a *12-13% decrease in aboveground biomass in both Exp 1 (not statistically significant) and Exp 3 (statistically significant). The shared findings across these different experimental contexts help to strengthen our results and suggest that while several grass species are negatively impacted by heatwaves, strategically choosing the least sensitive species in restoration could help lessen the impact of the heatwave on the vegetation-fire feedback loop.

Interestingly, in Exp 2, we also found that the occurrence of a heatwave decreased belowground biomass in S. nutans (25% reduction), in addition to the observed decrease in aboveground production (16% reduction). While this decrease belowground would not directly impact the vegetation-fire feedback loop, it has an indirect impact by decreasing the ability of grasses to regrow following fire the following summer. In fire prone ecosystems that experience lowseverity fires, dominant grass species consistently lose aboveground biomass through burning, while belowground root systems remain intact (Blair et al. 2014 ). These belowground roots and storage organs (e.g., rhizomes and tubers), which carry over from year to year, serve as a regenerative source of carbon for these dominant grasses to utilize in aboveground biomass production following fire (Bowen and Pate 1993) ; (Bellingham and Sparrow 2000) . When belowground reserves are depleted, growth the following year can be dampened, again leading to low fuel production, and hindering the successful establishment of the vegetation-fire feedback loop in subsequent year. Additionally, the decrease in belowground production has consequences for carbon sequestration. In general, terrestrial plant productivity is a major carbon sink, capturing carbon from the atmosphere and sequestering it in plant tissues, both above-and belowground . Grasslands and savannas store much of their carbon belowground, especially those that experience frequent fire events (Johnson and Matchett 2001) ; (Kitchen et al. 2009 ). Thus, grassdominated systems are important for global carbon sequestration (Scurlock and Hall 1998) . As such, carbon sequestration is an important ecosystem service provided by LLP savannas and is frequently a target of restoration efforts. Our work with Exp 2 further reinforces the need to understand the impact of heatwaves both above and belowground and suggests that this might be a particularly sensitive consequence of increased heatwave occurrence.

Our findings generally align with the results of previous work from other grass-dominated ecosystems. At a tallgrass prairie site, aboveground production of dominant species significantly decreased due to increased heat (infrared heat lamps) and decreased precipitation (drought-simulating rain-out shelters) in situ (Hoover et al. 2014) . Further, belowground production in grasses under heatwaves in a Mongolian grassland decreased, which affects grass resilience (Li et al. 2020) . Interestingly though, when isolating the effects of heat and water stress in a tallgrass prairie, drought (reduced precipitation) significantly reduced photosynthesis and productivity in A. gerardii and S. nutans, but the combination of heat and water loss (what our study calls a heatwave) did not add to any additional negative responses. This suggests drought alone dictates plant responses to heatwave stress (Hoover et al. 2014) . While the experiments presented here cannot tease apart the interactive effects of drought and heat, we plan to explore this idea in the future through further greenhouse studies.

In this study, we saw that a single heatwave event caused a significant decrease in biomass. Practically, this means that there will be less forage and cover for animals, as well as less fuel for the even spread of fire, thereby lessening the potential for successful restoration. The structure of these species-rich LLP savannas is maintained by frequent fire that is carried across the understory, primarily by bunchgrass species. These greenhouse studies indicate that future climate change may disrupt ecosystem processes in newly restored LLP savannas because the spread of fire is greatly impacted by a decrease in grass productivity (Brockway and Lewis 1997) . From these studies, we have seen that with an increased frequency of heatwaves, in field situations, it is likely that each grass plug, regardless of species, will be smaller and contribute less biomass to the system. With this information, land managers can be prepared to restore the understory plant community with greater efficacy. This loss of aboveground biomass can be overcome by planting resilient grass species, as well as a greater density of grass plugs, to maintain the desired amount of productivity post heatwave events. However, too much grass biomass can lead to a decrease in plant biodiversity, as large, abundant grasses assume dominance in the understory. It is therefore important to determine precisely how many more grass plugs to plant. In these studies, we found A. stricta to be more resilient than other species, with the smallest biomass reduction (12%) after a heatwave. Restoration plans could therefore for example potentially incorporate a 12% increase in the number of A. stricta grass plugs that are planted to overcome losses from heatwaves without reducing plant biodiversity.

While the findings of this study have important implications for the management of LLP savannas in a changing environment, it is important to note a key limitation of this study. Primarily, all three experiments were conducted under greenhouse conditions and therefore do not necessarily represent the effects of natural heatwaves on understory grass species in their native environment. In our experiment, grass plugs were planted in their own, individual pots, free from belowground competition; in the field, however, there would be competition between understory and the overstory species for shared resources, such as soil water and light availability (Pecot et al. 2007 ). Light availability, as a result of LLP tree density, greatly affects the composition and abundance of herbaceous understory species (Harrington et al. 2003) ; (Harrington et al. 2006 ). However, while competition between understory and overstory species may be of importance, in this species-rich community, herbaceous understory competition is relatively low (Roth et al. 2008) . Grass species native to LLP savannas have extensive root systems, though not as deep as their longleaf pine tree counterparts (Clewell 1989) ; (Nippert et al. 2012) , that grow well in sandy soils. Understory grass species in the field could develop deeper roots, accessing deeper resources, and therefore may show muted impacts of heatwaves compared to grasses grown in more restrictive pots. However, in the first year of growth from a plug in the field, it is unlikely the grass plugs would be able to develop the substantial rooting system seen in these perennial grasses in native LLP savanna; therefore, our greenhouse experiment is informative of first year restoration efforts (Tony et al. 2014); (Wonkka et al. 2017) . While field confirmation of our findings is still needed, these results assess the sensitivities of morphological traits that are behind potential decreases in aboveground biomass during a heatwave in the first year.

Additionally, the effects of a single heatwave event in the first year of restoration could be compounded by subsequent heatwave events. Further, understory grass biomass accumulation is essential for even fire spread, and, even if the grass plugs used in restoration are only temporarily impacted by a single heatwave, but decrease fuel load (biomass), fire effectiveness can be hindered (Addington et al. 2015) ; (Gonzalez-Benecke et al. 2015) . Without efficient and even fire, woody species can encroach, outcompete the herbaceous understory, and change the composition of the savanna (Kush et al. 1999 (Kush et al. , 2000 . Thus, even if the impacts of a single heatwave only seem temporary, the ramifications for the trajectory of the restoration process could be longer term, especially since multiple heatwave events in a single growing season are increasing in frequency (Ades et al. 2020) . Therefore, in future greenhouse studies, we plan to address the legacy effects of heatwaves on these dominant understory grasses for experiencing a subsequent heatwave event.

Overall, this study will help to ensure the successful restoration of the LLP savanna ecosystem in the face of a rapidly changing world in two key ways. First, these results provide valuable information to restoration practitioners. Heatwaves are already more prevalent and longer lasting than they historically were, and this trend is predicted to intensify (Overpeck 2013); (Mazdiyasni and AghaKouchak 2015) ; (Ades et al. 2020) . Our findings highlight the need for restoration practitioners to consider the added stress of heatwaves when making restoration plans, potentially selecting for more heatwave tolerant species, and increasing plug numbers. Restoring the vegetation-fire feedback loop may require different restoration methods in the future than historically implemented and may require that land managers quickly change restoration plans with the most up to date research and climate models informing plans (Clark et al. 2018 ). Second, the use of undergraduate research in the classroom setting promises to aid in restoration efforts. Not only does the CURE expose students to research and science (Russell et al. 2007 ); (Thiry and Laursen 2011), ensuring a diverse and scientifically literate workforce (Barlow and Villarejo 2004) ; (Eagan et al. 2011) , it also creates an emotional tie and a sense of pride and stewardship in * 250 students per year for this local and endangered ecosystem. These undergraduates conducted the experiments, collected and analyzed data for a final report, and gave group presentations to the class on their findings. Importantly, each semester the students also virtually attend a conference where this work is presented, and watch presentations of previous semesters work ensuring that they understand these data are needed and valued by the longleaf pine restoration community. Continuation of this CURE will allow us to educate and expose potentially thousands of scientists to this imperiled ecosystem and to the need to prioritize LLP savanna restoration. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Restoring longleaf pine: effects of seasonal prescribed fire and Overstory density on vegetation structure of a young longleaf pine plantation

Prescribed fire effects in a longleaf pine (Pinus palustris) ecosystem: Are winter fires working?

Making a difference for minorities: Evaluation of an educational enrichment program

Longleaf pine ecosystem restoration: the role of fire

Modelling responses of pine savannas to climate change and large-scale disturbance

Resprouting as a life history strategy in woody plant communities

Controlling the False Discovery Rate: A Practical and Powerful Approach to

Multiple Testing

Grassland Ecology

The Significance of Root Starch in Postfire Shoot Recovery of the Resprouter Stirlingia latifolia R. Br. (Proteaceae)

Long-term effects of dormantseason prescribed fire on plant community diversity, structure and productivity in a longleaf pine wiregrass ecosystem

Longleaf Pine Regeneration Ecology and Methods

Strong legacy of agricultural land use on soils and understory plant communities in longleaf pine woodlands

Human Dimensions of Ecological Restoration: Integrating Science, Nature, and Culture

Evaluating Climate Change Planning for Longleaf Pine Ecosystems in the Southeast United States

Natural History of Wiregrass (Aristida strida Michx., Gramineae)

Flammability of the keystone savanna bunchgrass Aristida stricta

History and Future of the Longleaf Pine Ecosystem

Fuels and fires influence vegetation via above-and belowground pathways in a high-diversity plant community

Understory plant biomass dynamics of prescribed burned Pinus palustris stands

Plant competition, facilitation, and other overstory-understory interactions in longleaf pine ecosystems

Above-and belowground competition from longleaf pine plantations limits performance of reintroduced herbaceous species

Ecological Restoration and Global Climate Change

Our Changing Climate. Impacts, Risks, and Adaptation in the United States: The Fourth National Climate Assessment

The longleaf pine forest: Long-term monitoring and restoration of a management dependent ecosystem

Contrasting sensitivities of two dominant C4 grasses to heat waves and drought

The effects of climate change on natural ecosystems of the Southeast USA

Spacing and competition between planted grass plugs and pre-existing perennial grasses in a restoration site in

Biodiversity increases the resistance of ecosystem productivity to climate extremes

Fire and grazing regulate belowground processes in tallgrass prairie

Annual fire and mowing alter biomass, depth distribution, and C and N content of roots and soil in tallgrass prairie

Patterns and Controls of Terrestrial Primary Production in a Changing World

Effects of extreme weather events on plant productivity and tissue die-back are modified by community composition

Understory plant community response after 23 years of hardwood control treatments in natural long leaf pine

Understory Plant Community Response to Season of Burn in Natural Longleaf Pine Forests

Fire and forest ecology: innovative silviculture and vegetation management Tall Timbers Fire Ecology Conference Proceedings

Ecological Restoration in the Face of Global Climate Change: Obstacles and Initiatives

Drought and heat wave impacts on grassland carbon cycling across hierarchical levels

Substantial increase in concurrent droughts and heatwaves in the United States

Root characteristics of C4 grasses limit reliance on deep soil water in tallgrass prairie

How global biodiversity hotspots may go unrecognized: lessons from the North American Coastal Plain

History and current condition of longleaf pine in the southern United States. General Technical Report -Southern Research Station

Competitive responses of seedlings and understory plants in longleaf pine woodlands: separating canopy influences above and below ground

Changes in regional heatwave characteristics as a function of increasing global temperature

The combined effects of an extreme heatwave and wildfire on tallgrass prairie vegetation

How important is competition in a species-rich grassland? A two-year removal experiment in a pine savanna

THE PIPELINE: benefits of undergraduate research experiences

The global carbon sink: a grassland perspective

Restoring native grasses in California Old Fields

The role of student-advisor interactions in apprenticing undergraduate researchers into a scientific community of practice

Incorporating plant mortality and recruitment into rangeland management and assessment

Fire frequency, agricultural history and the multivariate control of pine savanna understorey plant diversity

Toward an oldgrowth concept for grasslands, savannas, and woodlands

Composition and species diversity of pine-wiregrass savannas of the Green Swamp, North Carolina

Restoring the Ground Layer of Longleaf Pine Ecosystems

Acknowledgements We gratefully acknowledge the support of the CUR Transformations project, which is supported by the National Science Foundation (NSF) through an NSF DUE IUSE grant to the Council on Undergraduate Research (#16-25354) as