Water Dynamics in the Understory of a Pine Plantation Forest After Variable Retention Harvesting

Background: Variable Retention Harvesting (VRH) is a silvicultural technique applied to enhance forest growth, and restore forest stands to closely resemble their natural compositions. This study used sapow and understory eddy covariance ux measurements to examine the impacts of four different VRH treatments on the dominant components of evapotranspiration including canopy transpiration and water ux from understory vegetation and soil. These VRH treatments were applied to an 83-year-old red pine (Pinus resinosa) plantation forest in the Great Lakes region in Canada and included 55% aggregated crown retention (55A), 55% dispersed crown retention (55D), 33% aggregated crown retention (33A), 33% dispersed crown retention (33D) and unharvested control (CN) plot. Results: Study results showed a positive relationship between thinning intensity and the growth of understory vegetation, and hence enhanced evapotranspiration. The contribution to evapotranspiration from understory vegetation and soil was more pronounced in the dispersed thinning treatments, as compared to the aggregated. Overall, canopy transpiration contributed to 83% of total evapotranspiration in the un-thinned control plot and 55, 58, 30, and 23% for the 55A, 55D, 33A and 33D plots, respectively. The thinning or retention harvesting enhanced the water use eciency in all treatments. Conclusion: Our results suggest VRH treatments that follow a dispersed harvesting pattern may provide the optimal balance between forest productivity and evapotranspiration or stand water use. Furthermore, a balance of contributions from both the canopy and successional understory vegetation and soil, as observed in the 55% retention harvesting treatment, may increase the resiliency of forest to climate change. These ndings will help researchers, forest managers and decision-makers to improve their understanding of thinning impacts on water and carbon exchanges in forest and adopt

a balance of contributions from both the canopy and successional understory vegetation and soil, as observed in the 55% retention harvesting treatment, may increase the resiliency of forest to climate change. These ndings will help researchers, forest managers and decision-makers to improve their understanding of thinning impacts on water and carbon exchanges in forest ecosystems and adopt appropriate forest management practices to enhance their carbon sequestration capabilities, water use e ciency and resilience to climate change.

Background
Forest ecosystems play a signi cant role in global water and carbon cycling through evapotranspiration (ET) and photosynthesis processes, respectively. It is estimated that approximately 61% of the 117,600 km 3 of annual global precipitation is derived from terrestrial ecosystems (Schlesinger and Jasechko 2014). Further, more than 50% of this atmospheric moisture originates as transpiration from plants, predominantly forests and crops (Jasechko et al. 2013, Wei et al. 2017, Sheil 2018). In the past century, land-use changes have increased at an alarming rate. Globally deforestation is removing 18.7 million acres of forest every year (FAO, 2016). It is estimated that 18% of current climate warming trends can be attributed to deforestation and land-use change (Ellison et al. 2017, Alkama and Cescatti 2016). One recent study estimated that due to these alterations to terrestrial land cover, there is about 5-6% reduction in atmospheric water at a global scale (Sterling et al. 2013). As the climate is changing and global forest cover is decreasing, it is becoming more important to understand the intricate processes that drive water and carbon cycling at the land-atmosphere boundary. There is a growing need for restoring forest ecosystems through various means such as afforestation and reforestation and developing sustainable forest management methods to enhance forest growth, promote carbon sequestration and sustain and secure regional water resources.
In Canada red pine (Pinus resinosa) is a major plantation species and over 70% of plantation forest in Ontario are comprised of red pine (Kim 2020). It is a favourable species due to the straight, robust trunk, resiliency to drought conditions and shade tolerance (Magruder et al. 2013, Sharma andParton 2018).
Red pine stands were widely planted in the early 20th century to convert abandoned agricultural lands to native forest ecosystems. The management of plantation stands has been a challenge and traditional silviculture techniques are often inadequate to enhance stand growth and productivity (Beese et al. 2019). Therefore, forest managers and planners are striving to explore different forest management techniques that can not only increase stand growth but also enhance carbon sequestration, water use e ciency, biodiversity and resilience to climate change.
Variable retention harvesting (VRH) is a selective-thinning silvicultural method designed to increase forest growth, promote productivity and increase carbon sequestration (Franklin et al. 1997 While micrometeorological techniques, such as eddy covariance (EC) are widely used to measure ET above forest ecosystems (Baldocchi 2003(Baldocchi , 2020, the use of EC systems below a forest canopy is far less common due to numerous challenges such as low wind speed, weak and intermittent turbulence and large surface heterogeneity (Baldocchi et al. 2000, Launiainen et al. 2005. Some studies, however, have successfully measured carbon and water uxes below the forest canopy and partitioned ET into soil evaporation (E S ) and transpiration (T C ) (Baldocchi and Vogel 1996, Black et al. 1996, Constantin et al. 1999, Mission et al. 2007, Brown et al. 2014). But none of these studies were conducted in forests where different management regimes have been applied to evaluate their effectiveness for stand growth, carbon sequestration and water conservation.

Page 4/24
The objectives of this study are to (i) measure ET in four different VRH treatments and a control plot in a red pine plantation forest in the Great Lakes region in Canada (ii) partition ET into canopy and understory components of water uxes in each plot (iii) determine the water use e ciency of both the canopy and the understory in each treatment and (iv) explore which of these VRH treatments might be best suited to enhance stand growth while conserving water resources. This study is among the rst efforts to study and partition ET into its components in different VRH treatments in pine forests.

Site Description
The study site is located within the St. Williams Conservation Reserve (SWCR, 42°42'N, 80°21'W), about 3 km north of Lake Erie in southern Ontario, Canada. The temperate forest stand is a 21-hectare red pine (Pinus resinosa) plantation forest established in 1931 and is further referred to as 'CA-TP31'. In 2014, the plantation underwent variable retention harvesting (VRH) to restore the coniferous monoculture to a native Carolinian composition. Soils in the region are well-drained, sandy loam with a low to moderate water holding capacity. CA-TP31 is part of the larger Turkey Point Observatory, which consists of three As part of the VRH scheme, CA-TP31 was segmented into 21 one-hectare blocks and randomly treated with one of 5 harvesting techniques that differed in harvesting density and pattern: 33% basal retention in a dispersed pattern (33D), 55% retention in a dispersed pattern (55D), 33% retention in an aggregated pattern (33A), 55% retention in an aggregated pattern (55A) and an unharvested control (CN). The aggregated pattern of harvesting left remaining trees in small and large groups ( Figure 1). Further details are given in Bodo and Arain (2021b).
Since the implementation of VRH, successional species have emerged in the understory of the harvested blocks, with varying degrees of growth. Species include black oak (Quercus velutina), red maple (Acer rubrum), black cherry (Prunus serotina), and white pine (Pinus strobus). There was almost no understory in control plots where the canopy was almost closed.

Understory Eddy Covariance Flux Measurements
Carbon (CO 2 ), latent (LE) and sensible heat (H) uxes were measured over the understory in each VRH treatment during the 2019 growing season using a roving open-path eddy covariance (OPEC) system. The OPEC system was installed in one block of each treatment, for a minimum of 14 days before rotating to the next block (Table 1). Data collected on the day in which the instrument was moved was not included in the analysis. The instruments were installed in the centre of the plot at 5 m above the ground. It consisted of an infrared gas analyzer (Li-7500, LI-COR Inc.) and a 3D sonic anemometer (CSAT3, Campbell Scienti c Inc.). Flux measurements were made at 20Hz and averaged every 30-minutes. Meteorological measurements such as photosynthetically active radiation (PAR; LI190SB, LI-COR Inc.), air temperature (T a ) and humidity (RH; HC2S3, Campbell Scienti c Inc.), soil temperature at depths of 5 and 10 cm below the ground surface (T s ; TS107b, Campbell Scienti c Inc.), volumetric water content at depths of 5 and 10 cm below the ground surface (θ; CS616, Campbell Scienti c Inc.) were sampled every 5 seconds, averaged every half-hour and stored on a data logger (CR5000, Campbell Scienti c Inc.). Net ecosystem exchange (NEE U ) was calculated as the sum of the vertical CO 2 ux and the rate of storage in the air column below the IRGA. NEP U was then calculated as the opposite of NEE U (multiplied by -1).
All meteorological and ux data were processed following Brodeur (2014). Meteorological and ux measurements were cleaned using a two-step process described in Beamesderfer et al. (2020). All halfhourly uxes were subjected to friction velocity (u*) ltering to remove values that may be underestimated during periods of low turbulence. We used the moving-point determination method (Reichstein et al. 2005) to estimate u* threshold values for the understory. The u* threshold value was 0.064 m s -1 and the resulting ux data recovery following threshold ltering was 62%. Finally, carbon and water ux measurements collected during rain events (precipitation > 0.5 mm in a half-hourly interval) were considered erroneous and discarded.

Above-canopy Eddy Covariance Flux Measurements
Above-canopy uxes were measured using a reference eddy covariance system (EC REF ) installed above the white pine forest stand (CA-TP39), situated about 1 km north of CA-TP31. This ux station was chosen as a reference system due to the similar stand age and density as that of CA-TP31. Sap ux density (J s ; gH 2 O m -2 s -1 ) was calculated following Granier (1987). Tree-level sap ux measurements were scaled to plot-level transpiration for each of the study blocks following equation 1: Where T is transpiration (mm s -1 ), i denotes the treatment plot, J s is the average sap ux density of all sensors in plot I (gH 2 O m -2 s -1 ), and A s /A g is the ratio of sapwood area to total wood area in the plot (m 2 m -2 ).
Water Use E ciency Canopy water use e ciency (WUE C ) was estimated as the ratio of net primary productivity (NPP C ) to canopy transpiration (T C ) for each of the ve treatment blocks (g C m -2 /Kg H 2 O). Tree-ring width analysis was used to estimate carbon uptake (NPP C ) in the red pine canopy of each treatment for the growing season. Tree cores were collected 1.3 metres off the ground surface using a 5-mm increment borer as

Meteorological Conditions
Meteorological conditions conformed to typical seasonal averages for the 2019 growing season when compared to the previous ve years. Both thinning intensity and pattern in uenced below canopy radiation ( Figure 2). For example, in the un-thinned control plot, only 8% of PAR reached the ground surface due to the dense canopy. By comparison, in the 33A and 33D plots, on average 26% and 36% of PAR reached the ground, respectively. In the 55A and 55D plots, 18% and 25% of PAR penetrated the canopy to reach the ground surface. These values also suggest that when compared to the aggregated plots, the dispersed pattern of thinning allows for slightly more radiation to penetrate the canopy, which may be an important factor for understory growth and productivity. During the day, Ta was cooler below the canopy in each of the treatment plots when compared to the above canopy reference Ta on top of EC tower at CA-TP39 as expected. At night, a temperature inversion was observed in each of the treatments where the below canopy Ta was higher than the above the forest. There was no correlation between VRH treatment and difference in air temperature between the above and below canopy sensors. Ts measurements taken at 5 cm and 10 cm depths closely followed Ta. The driest measurement period was that of the 55D plot (25 July to 13 August) where only 21 mm of precipitation fell over the 12-day period. By contrast, between 14 August and 30 September we observed 118 mm of rainfall, while the EC U was measuring uxes in the control plot.

Partitioning of Evapotranspiration
There is a positive relationship between the level of thinning and presence of understory vegetation, with the more heavily thinned blocks (33A and 33D) experiencing the most understory growth. By contrast, control plots had least understory with the dominant understory vegetation species mostly comprising the non-vascular bryophytes. These differences in understory vegetation among VRH treatments had signi cant impact on understory ET. We observed the largest understory ET uxes in the most heavily thinned VRH treatments (33A and 33D) and the lowest in the un-thinned control (Figure 3c). On average, the understory ET in the control plot represented 17% of total ET (ET U + T C ). In the moderately thinned 55A and 55D treatments, ET U contributed to 45% and 42% of total ET, respectively; and in the 33A and 33D, it contributed up to 70% and 77% of total ET, respectively (Figure 3a,c). Further, daytime ET values measured in the understory were linearly correlated with the reference above-canopy ET measurements Additionally, the dispersed VRH treatments (33D and 55D) exhibited greater contribution of ET from the understory when compared to the aggregated plots of the same thinning intensity (Figure 4).
We observed the opposite trend in plot-level transpiration, with an average of 83% of total ET (Tc + ET U ) in the un-thinned control plot comprised of T C . On some days during the study period, the T C / ET U ratio was as high as 1 in the CN plot ( Figure 5). In the 33D plot, however, we saw T C /ET values as low as 0.12, with an average ratio of 0.23. Plot-level transpiration closely re ected trends in the stand's tree-density among the VRH treatments.
When comparing the reference above canopy ET (ET REF ) with the measured ET (ET U + T C ), values came within 10% accuracy limit. In the CN and 33A plots, ET was overestimated, by an average of 5%, where in the 55A, 55D and 33D plots, ET was slightly underestimated by an average of 9%. Overall, daily average ET re ected ET REF , with the aforementioned components responsible for varying contributions (Figure 6).

Water Use E ciency
Canopy-level water use e ciency, WUE C (NPP C /T C ) followed the growth trends of treatment plots with CN < 33A < 55D < 55A < 33D in the 2019 growing season ( Table 2). The plots with largest net primary productivity (NPP) were 55D (515 g C m -2 ) and 33D (481 g C m -2 ). While plot-level productivity was among the highest in 33D, this treatment exhibited the least amount of transpiration (104 mm) during the growing season, therefore the WUE C was 4.63 g C m -2 per kg H 2 O -the highest of all treatments.
Conversely, in the un-thinned control, plot-level transpiration was highest (297 mm), partly due to the large stand density (432 trees ha -1 ). Growing season NPP in the CN plot was moderately low when compared to the other treatments (258 g C m -2 ) but transpired more water, which led to a very low WUE C of 0.87 g C m -2 per kg H 2 O.
In the understory, WUE U (GEP U /ET U ) followed the general trend with 55A < 55D < CN < 33D < 33A with slight differences among these values (Table 2). Due to measurements having been collected at different time periods during the growing season and for varying durations, we cannot compare understory gross ecosystem productivity (GEP U ) between treatments. However, the ratio of GEP U /ET U and therefore, WUE U is upheld regardless of timing and duration. Interestingly, we observed the highest WUE U in the most heavily thinned treatments, 33A and 33D where the WUE U was 1.32 and 1.27 g C m -2 per kg H 2 O, respectively; but the lowest WUE U among the 55A and 55D (1.03 g C m -2 per kg H 2 O in both treatments).

Effects of VRH treatments on Meteorological Conditions
Our study showed VRH treatments that follow a dispersed thinning method (33D and 55D) allow for more PAR to reach the understory. This is important for climate change mitigation as it may promote higher growth and productivity in understory vegetation, leading to an increase in carbon sequestration. In fact, Mission et al. (2007) found that the GEP of the understory may reach up to 39% of total canopy GEP and is highly in uenced by PAR that penetrates the canopy. While understory vegetation is in uenced by PAR, their study found leaf area index (LAI) was more closely linked to overall productivity of the understory and the water balance (Mission et al. 2007). Additionally, Mission et al. (2007) found daytime Ta was generally higher in the understory than above the canopy, in less-dense forests. Our study found the opposite was true, where daytime Ta was cooler beneath the canopy, due to shading provided by the remaining trees in all treatments. To better understand the effects of VRH on micrometeorological conditions, several measurements throughout the plot should be taken to account for spatial variation beneath the canopy. Additionally, the presence and abundance of understory vegetation may in uence advective ow and therefore, Ta

Partitioning Evapotranspiration
Our study found a signi cant positive relationship between VRH intensity and ET U driven by understory vegetation. Like ndings by Xu et al. (2020), we observed an increase in the contribution of ET U to total ET as a result of increased thinning intensity. Moreover, we observed greater understory contributions from the dispersed treatments (33D and 55D), suggesting this may be the preferred treatment pattern. These results follow similar trends in growth among remaining trees determined by Zugic et al. (2021) using tree-ring analysis in the same site (CA-TP31). Their study found higher growth in the dispersed treatments when compared to the aggregated plots of the same retention.
We also found a strong negative relationship between thinning intensity and the ratio of transpiration to total ET (T/ET). There have been several studies that have quanti ed the contribution of canopy transpiration to total ET at stand, national and global scales (Jasechko et  While there are relatively few studies that compare water balance components between thinning treatments, the importance of quantifying these contributions in forest ecosystems is widely accepted. Our study is the rst known study to quantify and partition evapotranspiration in red pine following VRH treatments. ET U is particularly important during periods of drought, when canopy transpiration is low due to stomatal closure (Simonin et al. 2007). Therefore, quantifying the contribution of the understory to ecosystem ET is key to predicting the effects of climate change on these forests, for determining the optimal management strategies and growth and survival of understory species contributing to richness of biodiversity.
Water Use E ciency  Livingston et al. (1999) found that the presence of understory vegetation led to an increase in WUE in Japanese red pine (Pinus densi ora) and white spruce (Picea glauca), respectively. More recent studies have also found thinning in Norway spruce leads to an increase in productivity related WUE (Gebhardt et al. 2014). These results support our ndings that despite a greater presence of understory vegetation, canopy thinning leads to more productivity, less plot-level transpiration and therefore a higher WUE C .
When compared to WUE C , WUE U was lower in all treatments except for the un-thinned control.

Conclusion
This study quanti ed the in uence of different forest management (variable retention harvesting) treatments (33 and 55% retention harvesting in dispersed and aggregate forms) on the partitioning of total evapotranspiration in a red pine plantation forest in the Great Lakes region, in Canada. We found a positive relationship between thinning intensity, understory vegetation, and therefore understory evapotranspiration. The contribution from understory vegetation was more pronounced in the dispersed thinning treatments, when compared to the aggregated. Additionally, we observed canopy transpiration contributed to 83% of total ET in the un-thinned control. Finally, we found that water use e ciency increased as a result of thinning in the remaining trees in all treatments. These ndings suggest variable retention harvesting in a dispersed pattern with 55% basal retention (more than half of the trees) may provide the optimal balance between forest productivity and evapotranspiration or water use. Furthermore, a balance of contributions from both the canopy and successional understory vegetation may increase forest resiliency to future threats associated with climate change such as droughts.
As forest ecosystems provide numerous ecosystem services such as wood products, carbon sequestration and clean and sustainable water resources, there is growing realization among forest managers and the scienti c community to develop and adopt forest management or silviculture techniques that balance wood production and ecosystem service. Therefore, our study will contribute to further advance these goals.

Consent for Publication
Not applicable.

Availability of Data and Material
The datasets used during this study are available from the authors upon request.

Competing Interests
The authors declare that they have no competing interests. Authors' Contributions AVB collected, cleaned and processed sap ow, meteorological data and eddy covariance ux data. AVB was a major contributor in writing the manuscript. AVB and MAA designed the experiment with grants received by MAA. All authors read and approved the nal manuscript.  Figure 1 Aerial photograph of the VRH plots at CA-TP31 from Google Maps (2016).

Figure 2
Half-hourly values of a) photosynthetically active radiation (PAR) from above the forest canopy (black) and below (red), b) air temperature (Ta), c) soil temperature (Ts) measured 5 cm (solid line) and 10 cm (dotted line) below the surface, d) volumetric water content (θ) measured 5 cm (solid line) and 10 cm (dotted line) below the surface and precipitation. The vertical dashed lines indicate the day at which the meteorological instruments were moved to the next plot.

Figure 3
Total daily evapotranspiration (ET) measured from the reference above canopy eddy covariance system at TP39 (a); total daily canopy transpiration measured using sap ow sensors in dominant red pine trees in CA-TP31 (b); and total daily evapotranspiration measured from the roving understory eddy covariance system at CA-TP31 (c). The vertical dashed lines indicate the date at which the understory eddy covariance system was moved to the next plot.

Figure 4
Relationship between hourly evapotranspiration (ET) measured above canopy at TP39 site (ETREF) and below the canopy (ETU) in each of the VRH plots.
Page 23/24 Figure 5 The ratio of canopy transpiration, Tc to total evapotranspiration, ET (Tc + ETU) measured in each of VRH plots. The vertical black line shows the average daily T/ET value during the study period, and the grey bar shows the range of daily values.